IRP Discussion Paper
No. 1420-14
Food Assistance During and After the Great Recession in Metropolitan Detroit
Scott W. Allard (Co-Primary Investigator) University of Chicago [email protected]
Sandra K. Danziger (Co-Primary Investigator)
University of Michigan [email protected]
Maria Wathen
University of Michigan [email protected]
Final Report to the IRP RIDGE Center for National Food and Nutrition Assistance Research Grants Program
Project Period: July 1, 2012 to January 30, 2014
April 2014 This research was funded through the Institute for Research on Poverty (IRP) RIDGE Center for National Food and Nutrition Assistance Research, which is supported by the Food Assistance and Nutrition Research Program (FANRP) of USDA’s Economic Research Service (ERS). The views and opinions expressed herein are those of the author and do not necessarily reflect those of the Institute for Research on Poverty or the Food Assistance and Nutrition Research Program, the Economic Research Service, or the United States Department of Agriculture. IRP Publications (discussion papers, special reports, Fast Focus, and the newsletter Focus) are available on the Internet. The IRP Web site can be accessed at the following address: http://www.irp.wisc.edu.
Abstract
Economic shocks produced by the Great Recession have contributed to rising food insecurity,
with 14.7 percent of U.S. households being food insecure in 2009, compared to 11.1 percent in 2007. At
the same time, SNAP caseloads increased by nearly 60 percent since 2007 and the program now reaches
more than 40 million persons. Using data from the first two waves of the Michigan Recession and
Recovery Survey (MRRS), a unique panel survey of a representative sample of working-age adults in the
Detroit Metropolitan Area, this project explores three research questions related to the receipt of SNAP
among low-income households: How have low-income families in the Detroit Metropolitan Area bundled
SNAP with other types of public assistance and help from charitable nonprofits in the wake of the Great
Recession? When controlling for economic shocks and respondent characteristics, to what extent is access
to local food assistance resources related to receipt of SNAP and charitable nonprofit food assistance?
How are receipt of SNAP assistance and economic shocks related to household food shopping behaviors
and food security? Several important findings emerge from this project that should be of interest to
scholars, policymakers, and advocates. Among them is the finding that food insecurity is quite prevalent
among poor and near-poor households in metro Detroit following the Great Recession. Fifty-one percent
of households below the federal poverty line were food insecure in the year prior to the Wave 1 survey.
Keywords: Food insecurity; Great Recession; SNAP; Michigan Recession and Recovery Survey
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Executive Summary
BACKGROUND AND METHODOLOGY
Economic shocks produced by the Great Recession have contributed to rising food insecurity,
with 14.7 percent of U.S. households being food insecure in 2009, compared to 11.1 percent in 2007. At
the same time, SNAP caseloads increased by nearly 60 percent since 2007 and the program now reaches
more than 40 million persons. Nonprofit food pantry use also increased during the Great Recession, with
an estimated 37 million persons using charitable food programs in 2009. Financial hardship has increased
as well in recent years due to the economic downturn, placing greater importance on understanding the
connections between food assistance, food security, and other types of material need.
Using data from the first two waves of the Michigan Recession and Recovery Survey (MRRS), a
unique panel survey of a representative sample of working-age adults in the Detroit Metropolitan Area,
this project explores three research questions related to the receipt of SNAP among low-income
households: How have low-income families in the Detroit Metropolitan Area bundled SNAP with other
types of public assistance and help from charitable nonprofits in the wake of the Great Recession? When
controlling for economic shocks and respondent characteristics, to what extent is access to local food
assistance resources related to receipt of SNAP and charitable nonprofit food assistance? How are receipt
of SNAP assistance and economic shocks related to household food shopping behaviors and food
security?
Data for this project comes from the first two waves of the Michigan Recession and Recovery
Survey (MRRS), which gathers detailed information about employment history, income sources, food
security, safety net program participation, private social support, material hardships, health and mental
health, grocery shopping habits, and basic household demographics from a representative sample of
households with adults aged 19 to 64 years living in the three-county Detroit metropolitan Area. Wave 1
of the MRRS completed hour-long in-person interviews between late October 2009 and March 2010 with
914 adults between the ages of 19 and 64 (response rate of 82.8 percent). The second wave (also hour-
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
long in-person interviews) was completed between April and August 2011 with 847 respondents
(response rate of 93.9 percent). Information about the residential location of each MRRS respondent is
used to assess household proximity and accessibility to a number of different food assistance and retail
resources: SNAP administrative offices; food pantries; SNAP authorized retailers; and food retailers as
reported by InfoUSA marketing data. In doing so, this study is in a unique position to connect household-
level food outcomes (i.e., program participation, food security, grocery shopping habits) to the local food
resource infrastructure with a precision not found in most published food policy research.
FINDINGS
Several important findings emerge from this project that should be of interest to scholars,
policymakers, and advocates. First, food insecurity is quite prevalent among poor and near-poor
households in metro Detroit following the Great Recession. Fifty-one percent of households below the
federal poverty line were food insecure in the year prior to the Wave 1 survey. Consistent with recent
research findings that food insecurity is more prevalent among non-poor populations than is commonly
realized, 36.3 percent of households with income between 100 to 200 percent of the federal poverty line
and 33.4 percent of households with income between 200 to 300 percent of federal poverty were
classified as food insecure in the twelve months prior to Wave 1. Perhaps reflecting effects of the
economic recovery, the prevalence of food insecurity declined slightly between waves of the MRRS.
Nevertheless, more than 40 percent of poor households and about one-third of near-poor households were
food insecure in Wave 2.
Nearly 70 percent of Wave 1 households with income below the poverty line report receiving
SNAP benefits at some point in the prior year, compared to 27.9 percent of households with income
between 100 and 200 percent of federal poverty. Receipt of charitable food assistance is less common
among poor and near-poor households in Detroit than is receipt of SNAP. About one-third of poor
households report receiving charitable food assistance and less than 15 percent of households between
100 percent and 200 percent of poverty report receiving help from nonprofit charities. As might be
ii
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
expected, we find that households experiencing periods of unemployment and detachment from the labor
force due to work limiting health conditions are more likely to receive public and private food assistance
than households that do not experience spells of unemployment or detachment from the labor force.
This study also finds evidence that low-income households receiving SNAP reside closer on
average to both SNAP administrative offices and food pantries than low-income households that do not
receive SNAP. There also is some evidence that poor households receiving SNAP are located slightly
closer to SNAP retailers and chain grocery stores than poor households not receiving SNAP. In contrast
to expectations from some of the existing literature on food deserts, however, we find that poor
households in Detroit on average are slightly closer to the nearest SNAP retailer, SNAP grocery store, or
chain grocery store. Differences in spatial access to food assistance program resources (e.g., SNAP
offices or food pantries) is found to be associated with increased likelihood that a household receives
assistance, even after controlling for relevant demographic and economic factors. For example, being
within 1 mile of a SNAP administrative office increases the likelihood of receiving SNAP assistance by
about one-third over the baseline case where the household is more than a mile away from a SNAP office.
Contact: Scott W. Allard (Co-Primary Investigator) Associate Professor, School of Social Service Administration, University of Chicago 969 E. 60th Street, Chicago, IL 60637 Phone: (773) 702-1131 Email: [email protected] Sandra K. Danziger (Co-Primary Investigator) Professor, School of Social Work, University of Michigan 735 South State Street, Ann Arbor, MI 48109 Phone: (734) 615-4648 Email: [email protected]
iii
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Food Assistance During and After the Great Recession in Metropolitan Detroit
INTRODUCTION1
While the Great Recession officially ended in June 2009, rates of food insecurity and receipt of
food assistance persist above their pre-recession levels. From 2008 to 2011, over 14 percent of
households were food insecure at some time during the year, compared to about 11 percent of households
from 1999 to 2007 (Coleman-Jensen, Nord, Andrews, and Carlson, 2012). SNAP participation rates and
use of emergency food assistance programs similarly increased in the last five years. Between December
2007 and October 2010 the SNAP caseload increased by nearly 60 percent and reached more than 40
million persons during a typical month in 2010 (Center on Budget and Policy Priorities, 2010; Eslami,
Filion, and Strayer, 2011). Nonprofit food pantry use increased during the Great Recession and an
estimated 37 million individuals received help from charitable food programs in 2009, including a large
percentage of SNAP recipients (Mathematica Policy Research, 2010; U.S. Conference of Mayors, 2008).
Financial hardship has increased as well in recent years due to the economic downturn, placing greater
importance on understanding the connections between food assistance, food security, and other types of
material need (Nord and Golla, 2009; Pan and Jensen, 2008; Shaefer and Gutierrez, 2012; Yen, Andrews,
Chen, and Eastwood, 2008).
At the same time, there has been a surge in interest around the impact of spatial context on the
presence, prevalence, and persistence of food insecurity. Much of the research to date has been focused
on the presence of “food deserts,” where limited spatial access to grocery stores or outlets of affordable
and fresh food is thought to be associated with lower household food security for adults and children.
1This research also received support from the Office of the Assistant Secretary of Planning and Evaluation, U. S. Department of Health and Human Services, the Office of the Vice President for Research at the University of Michigan, the Ford Foundation and the John D. and Catherine T. MacArthur Foundation. Allard’s work on this project was supported by the Population Research Center at NORC and the University of Chicago, and National Institute of Child Health and Human Development (NICHD) Grant #5R24HD051152-07. The views and opinions expressed herein are those of the authors and do not necessarily reflect those of any of our project funders. Finally, we thank Sarah Burgard, Sheldon Danziger, Tedi Engler, Italo Gutierrez, Colleen Heflin, Chieko Maene, and Luke Shaefer for their assistance and comments on the project.
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Other aspects of place matter as well. For instance, some evidence suggests that the presence of nonprofit
food assistance programs also can vary widely by neighborhood and across communities, ironically being
less accessible to low-income populations most in need (Allard, 2009b).
Combined, these trends place greater importance on understanding the connections between food
assistance, food resource access, and food security (Nord and Golla 2009; Pan and Jensen 2008; Shaefer
and Gutierrez 2012; Yen, Andrews, Chen, and Eastwood 2008). Identifying where low-income
households found help during the downturn and subsequent recovery will highlight which population
subgroups today’s safety net is most likely to reach and which subgroups may be better served. Inquiry
into the use of public and private assistance may be especially important in the current context, as the
Great Recession was the most severe test of the contemporary safety net’s capacity to respond to
persistent need. Improved understanding of the spatial antecedents of food assistance and food insecurity
also could translate into more efficient allocation of public program dollars, private capital,
entrepreneurial activity, and philanthropic resources.
Using data from the first two waves of the Michigan Recession and Recovery Survey (MRRS), a
unique panel survey of a representative sample of working-age adults in the Detroit Metropolitan Area,
this IRP RIDGE Center for National Food and Nutrition Assistance Research grants program-supported
project explores several research questions related to the receipt of SNAP among low-income households:
How have low-income families in the Detroit Metropolitan Area bundled SNAP with other types of
public assistance and help from charitable nonprofits in the wake of the Great Recession? When
controlling for economic shocks and respondent characteristics, to what extent is spatial access to local
food assistance resources related to receipt of SNAP and charitable nonprofit food assistance? How are
receipt of SNAP assistance and economic shocks related to household food shopping behaviors and food
security?
Our findings to date provide insight into how low-income families combine governmental and
nongovernmental food assistance with other safety net programs to weather the effects of job loss and
economic recession. Such insights could prove useful for program outreach and efforts to enroll 2
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
households eligible for different types of public assistance. Our results further highlight how the spatial
distribution of food assistance resources may affect program take-up and indirectly shape food behaviors
among low-income households. Although nested within the Detroit metropolitan area, we think these
findings will help planning and coordination efforts in a variety of urban and suburban locations.
UNDERSTANDING RECEIPT OF FOOD ASSISTANCE IN THE CONTEXT OF THE GREAT RECESSION
The Great Recession had a dramatic and sustained impact on work, earnings, and poverty in most
communities in the United States. Decreases in work activity and median household income following the
end of the Great Recession were far more severe than any other recession since 1970. Increases in poverty
following the Great Recession were much higher than any other recession since 1980. The poverty rate
increased by two percentage points during the recession from 2007 to 2009, and then continued to
increase by almost a full percentage point in 2010 to peak at 15.1 percent nationally. During that three-
year span, nearly 10 million Americans fell below the federal poverty line (DeNavas-Walt, Proctor, and
Smith, 2011; DeNavas-Walt, Proctor, and Smith, 2012). Also in contrast to past recessions, the impact of
the Great Recession was felt in suburban, as well as urban and rural, communities.2
Consistent with the severity of the economic downturn, participation rates for most public
assistance programs increased directly in response to the Great Recession. According to a recent Census
Bureau report, 17.8 percent of the U.S. population—roughly 42 million persons—received assistance
from Temporary Assistance for Needy Families (TANF); General Assistance; Supplemental Nutrition
Assistance Program (SNAP, formerly food stamps); Supplemental Security Income (SSI); Medicaid; or
housing assistance in 2005. In 2009, about 45 million people (18.6 percent of the population) received
assistance from these public programs, an increase of about 6 percent (Kim, Irving, and Loveless, 2012).
Combined, public cash and in-kind safety net programs such as TANF, SNAP, the Earned Income Tax
2In fact, the initial effects of the recession were more severe in suburban areas than in the urban centers of the largest metropolitan areas in the United States, see Garr (2011).
3
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Credit (EITC), Medicaid, and Unemployment Insurance (UI) help delivered more than $300 billion in
benefits to tens of millions of low-income households in 2009 (Center on Budget and Policy
Priorities, 2012; Isaacs, Vericker, Macomber, and Kent, 2009; Kneebone, 2009; Simms, 2008; U.S.
Department of Health and Human Services, 2010).
Even though the recession officially ended in 2009, the effects of the downturn persisted for
many low-income households whose work opportunities and earnings have not returned to pre-recession
levels. Many safety net programs continued to expand, therefore, in the several years following the end of
the recession. Between December 2007 and April 2012, the SNAP caseload increased by about 70 percent
to serve more than 46 million persons, with the steepest growth in the program occurring between 2009
and 2011 (Center on Budget and Policy Priorities, 2012; Eslami, Filion, and Strayer, 2011). Medicaid
expenditures and enrollment have increased steadily since 2007 with the program serving 10 million more
individuals in 2011 than four years before (Kaiser Commission of Medicaid and the Uninsured, 2012;
Smith, Gifford, Ellis, Rudowitz, and Snyder, 2012). SSI caseloads continued to steadily increase
following the Great Recession, with the total caseload rising from about 7 million in 2007 to about 8
million individuals in 2012 (Congressional Budget Office, 2007, 2012; Schmidt, 2012). Similarly, per
capita expenditures for the EITC continued to increase in the few years following the recession (Moffitt,
2012).
As suggested by the discussion above, SNAP is one of the largest public assistance programs in
place today and is the largest public food assistance program in operation. SNAP provides monthly food
assistance benefits to households with gross monthly income at or below 130 percent of the federal
poverty level. Food stamps caseloads fell in the 1990s due to historic economic growth and welfare
reform, but SNAP caseloads have risen past pre-welfare reform levels since 2000. Particularly sharp
increases in SNAP participation occurred in the years during and after the Great Recession. Nationally,
SNAP caseloads increased by 69 percent from 2007 to 2012 and the program reached nearly 45 million
individuals in 2012 (Klerman and Danielson, 2012; U.S. Department of Agriculture, 2012a). The average
4
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
SNAP recipient received about $133 a month in benefits in 2012 and SNAP program expenditures in that
year reached about $81 billion (Center on Budget and Policy Priorities, 2013).
Complementing SNAP and other public food assistance programs are private nonprofit charities
and social service organizations that provide food and meals to low-income families in need. Nonprofit
food pantry use has increased since the Great Recession and an estimated 37 million individuals received
help from charitable food programs in 2009, including a large percentage of SNAP recipients (Feeding
America, 2011; Mathematica Policy Research, 2010). About one-third of food pantry clients received
help from a program at least once every month in a calendar year (Feeding America, 2011). Food pantry
use is more prevalent in cities and rural places, and in the South—areas where poverty rates tend to be
higher and families are at greater risk of not having enough food (Nord, Andrews, and Carlson, 2008).
SNAP assistance and other types of food assistance are expected to reduce hunger and improve
household food security. Food assistance, however, simultaneously seeks to improve food security and
allow households greater flexibility to reallocate income to other pressing needs or bills. In-kind
assistance programs like SNAP are intended to reduce household financial hardship by increasing
household purchasing power. Increasingly food assistance may serve as a gateway to other types of public
and private assistance. Although not a stated goal in many instances, therefore, receipt of food assistance
also may build awareness or connection to other programs of assistance for which low-income households
may be eligible.
Factors Associated with Receipt of Food Assistance
The existing research literature points to many factors that may shape how households draw upon
public or private sources of support and whether they bundle many different types of assistance over the
course of a year. Household economic circumstances are critical determinants of program participation.
Income poverty, job loss, decreased work earnings, and periods of unemployment or detachment from the
labor force lead to financial and material hardships. Families in poverty are more likely to seek public and
private assistance to address such hardships than households above the poverty line who also may be 5
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
experiencing such hardships.3 Households below and just above the federal poverty line are eligible for
many types of public assistance, but income eligibility varies by program. For example, SNAP reaches
households with income at or below 130 percent of the federal poverty line, but Medicaid can reach
uninsured adults and children above 150 percent of poverty. Private charities similarly consider poverty
status or income when delivering programs of support, although income eligibility rules often are more
flexible than those in public programs (Johnson, 2012; Kim, Irving, and Lovelace, 2012).
Several demographic factors are thought to matter as well. Female-headed households, who are
more likely to be poor and qualify for public programs than married couples or single male-headed
households, are much more likely to participate in means-tested public programs (Kim, Irving, and
Lovelace, 2012). Individuals without a high school or college degree may be more likely to receive public
assistance because of the difficulty finding a good paying job.4 Older women who head households and
those in good health have been found less likely to receive welfare cash assistance, food stamps, or public
housing assistance than younger women or those with poor health conditions (Keane and Moffitt, 1998;
Ratcliffe, McKernan, and Finegold, 2008). Participation rates in public programs are much higher among
blacks and Hispanics than whites.5 Race and ethnic differences in program participation may reflect lower
income and higher incidence of poverty among race and ethnic minorities. Legal restrictions on
immigrant access to public assistance programs, as well as gaps in cultural competency, and race or ethnic
discrimination lead to lower take-up of government cash and in-kind benefit programs.6 In addition, there
3Nearly three-quarters of persons with income below the federal poverty line received some type of public assistance in 2009, see Kim, Irving, and Lovelace (2012).
4Participation in means-tested public programs has been found to be twice as high among individuals without a high school degree, as among those with a high school degree (33.1 percent versus 17.8 percent respectively), see Kim, Irving, and Loveless (2012). Keane and Moffitt (1998) similarly find higher levels of education related to lower rates of welfare, food stamp, and housing assistance receipt.
5Kim, Irving, and Loveless (2012) find that 46.3 percent of blacks and 44.1 percent of Hispanics received assistance from a means-tested public program at some time in 2009, compared to 16.9 percent of whites.
6For instance, children from low-income immigrant households are roughly half as likely to access SNAP (or food stamp) benefits or TANF assistance as children from low-income native-born households, see Chaudry and
6
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
is evidence that predominantly black and Hispanic neighborhoods have less access to government and
nonprofit social service offices than predominately white neighborhoods (Allard, 2008; Allard and Roth,
2014).
A number of policy-related factors also shape program participation. Due to shared administrative
processes and categorical eligibility determinations, participation in certain public assistance programs is
highly correlated with participation in other public programs. For example, the vast majority of low-
income families receiving TANF, SSI, or SNAP benefits also are enrolled in Medicaid because those
programs often are coordinated in the same government offices or application procedures (Ratcliffe,
McKernan, and Finegold, 2008; Reese, 2006). As a result, more than 96 percent of TANF clients and 88
percent of SNAP clients in 2004 were receiving supports from at least two other public assistance
programs. On the other hand, programs like Unemployment Insurance often are administered through
government offices that do not provide other means-tested public programs. Thus, receipt of UI is only
weakly correlated with other means-tested programs (Reese, 2006). State-level variation in eligibility and
recertification policies can lead to state-level differences in program utilization. There is evidence that
states with more lenient food stamp eligibility rules or longer recertification periods see higher rates of
program participation than states with more restrictive eligibility or recertification policies (Ratcliffe,
McKernan, and Finegold, 2008). Additionally, there are complex interactions between benefit levels and
eligibility determinations that can dampen participation. Even though TANF and food stamp receipt are
highly correlated, stronger TANF sanction policies and TANF limitations on immigrant eligibility lower
rates of food stamp receipt among otherwise eligible individuals (Ratcliffe, McKernan, and Finegold,
2008; Fix, Capps, and Kaushal, 2009; Borjas, 2004).
Household characteristics of SNAP recipients provide insight into factors that may be associated
with food assistance program take-up. Roughly eight in ten households receiving SNAP have income
Fortuny (2010), Fix, Capps, and Kaushal (2009), Ratcliffe, McKernan, and Finegold (2008), and Soss, Fording, and Schram (2011).
7
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
below the federal poverty line. Families in extreme poverty—those in great need or risk—are more likely
to participate in the program.7 Households experiencing unemployment and those in areas with higher
unemployment rates are more likely to participate in the program.8 Controlling for other household
characteristics, studies find that blacks are more likely to receive SNAP than whites. Low levels of
completed education are found to be associated with higher likelihood of SNAP receipt. About half of all
households receiving SNAP contain children and slightly more than half of SNAP households with
children are single-parent households. Disability and health limitations also increase the likelihood that a
household receives SNAP (U.S. Department of Agriculture, 2009b; Hanratty, 2006; Klerman and
Danielson, 2011; Purtell, Gershoff, and Aber, 2012; U.S. Department of Agriculture, 2012a). Roughly
three of every four SNAP recipients are children, elderly adults, or individuals with a disability
(Congressional Budget Office, 2012).
Similar factors have been found to shape use of food pantries or other charitable emergency food
assistance programs. Racial minorities, single-parent households, individuals and households with low
levels of income, individuals in poor health, and low educational attainment have been found more likely
to receive food pantry assistance (Bartfeld, 2003; Bhattarai, Duffy, and Raymond, 2005; Feeding
America, 2011; Martin, Cook, Rogers, and Joseph, 2003). Many who use food pantries may lack access
to SNAP. For example, Algert, Reibel, and Renvall (2006) find that about 80 percent of food pantry
clients in two cities in Southern California received no SNAP benefits, despite having very low levels of
income. Low take-up of SNAP among these food pantry clients is associated with housing instability,
immigrant status, and lack of English language skills.
7Bartlett, Burstein, and Hamilton (2004) find that eligible families are less likely to believe they were eligible to participate in SNAP if they have income above the federal poverty line, had some assets (e.g., car or bank account), and were not experiencing food insecurity.
8A majority of SNAP households with a non-disabled working age adult are working while receiving benefits, but many recipients who are working are under-employed or have unstable attachment to the labor force (McKernan and Ratcliffe, 2003; Rosenbaum, 2013; U.S. Department of Agriculture, 2012b). Nevertheless, the percentage of recipients with work earnings has increased substantially over the past fifteen years (McKernan and Ratcliffe, 2003). Of working-age recipients that aren’t working, only about 40 percent were found to be not in the labor force, often because of a disability or caring for an adult or child with a disability (Rosenbaum, 2013).
8
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Apart from understanding which factors are associated with food assistance program
participation, it is important to assess how receipt of food assistance affects household food security and
other aspects of hardship due to recession, unemployment, or income poverty. Selection effects, however,
make it difficult to produce unbiased estimates of the association between food assistance receipt and
household food security because of unmeasured or unobserved characteristics likely correlated with both
food assistance receipt and household hardship (Daponte, Sanders, and Taylor, 1999; Gibson-Davis and
Foster, 2006; Gundersen and Oliveira, 2001; Nord and Golla, 2009). Many studies address selection by
using state-level food assistance policy variation to instrument for food assistance receipt. When using
instrumental variable approaches, these studies often find food assistance receipt to be associated with
lower risk of household food insecurity (Bartfeld and Dunifon, 2006; Shaefer and Gutierrez, 2012; Yen,
Andrews, Chen, and Eastwood, 2008).
Access to Food Assistance Resources and Food Retailers
Apart from standard household and individual characteristics commonly understood to shape
food assistance program participation, there is growing interest in degree to which spatial context affects
food assistance program participation rates and food security. Even though public assistance programs are
funded and regulated by the federal and state government, program application and administration often
occurs in local offices. While state-level program eligibility policies can affect the size of food assistance
program caseloads, such policies may be implemented less consistently or evenly between and within
local places (Bartlett, Burstein, and Hamilton, 2004; Klerman and Danielson, 2011; Soss, Fording, and
Schram, 2011). Moreover, nonprofit social service providers have discretion over what programs to offer,
which client populations to serve, and where to locate operations. Many factors constrain where public
and private nonprofit food assistance program offices are located, but chief among them can be
considerations about public transit accessibility, the cost of suitable office space, and the location of key
partners or funders (Allard, 2009b). Not all neighborhoods or communities will have easy access to public
9
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
or nonprofit providers, and the presence of such supports varies widely from place to place (Allard,
2009b; Allard and Roth, 2010).
Indeed, there is evidence that food assistance programs may not be as well-matched to the
location of need as might be imagined. Allard (2009b) finds high-poverty neighborhoods in metropolitan
Chicago, Los Angeles, and Washington, D.C., to have about 50 percent less access to emergency food
and cash assistance providers than low-poverty neighborhoods. Kissane (2010) underscores that spatial
access to community-based social service organizations, many of which offer emergency food assistance,
is critical to understanding which programs low-income households utilize. Interviews with low-income
women from the Kensington neighborhood in Philadelphia yielded evidence that even distances up to a
mile were too far for low-income households to manage. Interviews also underscore that perceived safety
and race or ethnic composition of the community, along with other aspects of social context, powerfully
shape which local organizations individuals feel comfortable to visit. In more suburban or rural areas, the
distances that clients and providers must travel to receive or deliver food assistance are higher and place
greater burdens on individuals or organizations. On top of these considerations, research has found rural
and suburban communities to have fewer, less well-resourced, and less accessible food assistance
providers than urban communities (Allard, 2009a; Allard and Roth, 2010).
There is evidence that greater proximity to safety net program providers will increase the
likelihood that low-income households will know about programs of assistance, receive referrals, and be
able to commute to those opportunities, which should translate into higher take-up of assistance (Allard,
2009b; Allard, Tolman, and Rosen, 2003; Bartlett, Burstein, and Hamilton, 2004; Kissane, 2003). For
example, SNAP clients may be expected to make re-certification visits and submit application materials
in-person (McKernan and Ratcliffe, 2003). Challenges finding child care and accessing administrative
offices during the workday are associated with lower SNAP take-up among eligible families (Widom and
Martinez, 2007). Distance from SNAP offices may increase time or commuting costs and thus discourage
participation (Bartlett, Burstein, and Hamilton, 2004; U.S. Department of Agriculture, 2010). Lack of
access to a car, lack of information about local programs, and difficulty carrying food home were the 10
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
most prominently cited reasons that low-income households in Hartford, CT, did not participate in local
food pantry assistance programs (Martin, Cook, Rogers, and Joseph, 2003).
Along with interest in the spatial dimension of food assistance resources, there has been a
significant rise in research examining the prevalence and location of “food deserts,” neighborhoods and
areas containing no or very few grocery stores or stores carrying fresh food items. The retail food
environment may be causally connected to food security because it shapes what households can purchase
and the extent to which the costs of food shopping are higher for residents of disadvantaged communities.
Closer proximity to food retailers, particularly those accepting SNAP benefits, may be related to food
assistance take-up as well. Not only may proximity to supermarkets or food stores accepting SNAP
benefits reduce time and transportation costs for food assistance recipients, but the presence of such stores
may also increase awareness of the program and its benefits. Access to supermarkets and chain stores
may be particularly important to food outcomes among food assistance recipients, as these types of stores
have been found to carry a wider array of fresh food items and to offer lower food prices than other types
of food retailers (U.S. Department of Agriculture, 2009a).
While the median U.S. household is 0.81 miles to the nearest supermarket and the average time
spent on travel to grocery shopping is about 15 minutes per day (U.S. Department of Agriculture, 2009a),
many studies find access to food retailers has been found to vary by race, ethnicity, and class composition
of a community. Studies often report that predominantly black and Hispanic neighborhoods have less
access to supermarkets and large chain grocery stores than predominately white areas. For example,
Gallagher (2006) finds that residents of majority African-American neighborhoods in Chicago have to
travel almost 40 percent farther on average to reach the nearest chain grocery store compared to residents
of majority white neighborhoods (0.77 miles versus 0.57 miles on average). Nationally, zip codes with
“higher proportions” of African-Americans have been found to contain half as many chain grocery stores
on average as zip codes with higher proportions of whites (Powell, Slater, Mirtcheva, Bao, and
Chaloupka, 2007). Lower income areas also have been found to contain fewer chain grocery stores than
middle or upper income areas (Powell et al., 2007; Moore and Diez Roux, 2006). A study of three 11
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
communities located in Maryland, New York, and North Carolina finds that “predominantly white” and
affluent census tracts contain twice as many supermarkets on average than predominantly black and
poorer areas when controlling for population size (Moore and Diez Roux, 2006). Similarly, Zenk and
colleagues (2005) find that high-poverty predominantly African-American census tracts in Detroit are
about 1.1 miles farther from the nearest chain supermarket compared to high-poverty predominantly
white tracts in Detroit.
Whether looking at spatial access to food assistance programs or food retailers, however, very
few studies can link the residential location of low-income individuals, food assistance recipients, or food
insecure households to the location of food resources with any real precision (Allard, 2014). Most studies
use census tracts or zip codes as the unit of analysis and focus on whether a program or store location falls
within those geographic boundaries, providing only blunt measures of access that risk misstating the
scope of gaps in food resource access. Most of the empirical work examining the relationship between
place, food assistance, and food security, therefore, emerges from studies that lack the conceptual and
empirical clarity required to assess local-level spatial processes.
RESEARCH QUESTIONS AND DATA
We propose to extend this literature in two ways. First, we propose to use the spatial variation in
the administration and availability of food assistance resources in Detroit to create a unique set of
instruments for modeling how SNAP program participation affects household outcomes in the Michigan
Recession and Recovery Survey (MRRS). Second, in addition to looking at food security and financial
hardship, we argue that studies of food assistance should look at changes in specific food consumption or
shopping behavior that would directly affect quality of diet or food security. A growing literature on food
deserts examines the presence of grocery stores and supermarkets in high poverty areas (Gallagher, 2006;
Haider and Bitler, 2009; Neckerman, Bader, Purciel, and Yousefzadeh, 2009; Sparks, Bania, and Leete,
2011), yet very little work has explored how food assistance is related to changes in where low-income
households shop.
12
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Using data from the first two waves of the Michigan Recession and Recovery Survey (MRRS), a
unique panel survey of a representative sample of working-age adults in the Detroit Metropolitan Area,
this IRP RIDGE Center for National Food and Nutrition Assistance Research Grants Program-supported
project explores several research questions related to the receipt of SNAP among low-income households:
How have low-income families in the Detroit Metropolitan Area bundled SNAP with other types of
public assistance and help from charitable nonprofits in the wake of the Great Recession? When
controlling for economic shocks and respondent characteristics, to what extent is spatial access to local
food assistance resources related to receipt of SNAP and charitable nonprofit food assistance? How are
receipt of SNAP assistance and economic shocks related to household food shopping behaviors and food
security?
To answer these questions, we use unique panel data from the Detroit Metropolitan Area during
2008 and 2010 to examine food security, food assistance, and coping responses to hardship among
households with income at or below 300 percent of the federal poverty line. Our data come from the first
two waves of the MRRS, which gathers detailed information about employment history, income sources,
education and training, safety net program participation, private social support, material hardships, health
and mental health, marital and relationship status, and basic household demographics from a
representative sample of households with adults aged 19 to 64 years living in the three-county Detroit
metropolitan area. Wave 1 of the MRRS completed hour-long in-person interviews between late October
2009 and March 2010 with 914 adults between the ages of 19 and 64 (response rate of 82.8 percent). The
second wave (also hour-long in-person interviews) was completed between April and August 2011 with
847 respondents (response rate of 93.9 percent).9 Demographic characteristics of MRRS respondents at or
below 300 percent of poverty can be found in Table 3 below.
9When survey weights are applied, the MRRS sums to the American Community Survey (ACS) estimated total population count for Macomb, Oakland, and Wayne counties of metropolitan Detroit; see Adams, Lepkowski, Elkasabi, and Battle (2011).
13
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
At each wave of the MRRS, respondents were asked a series of questions about food purchases
and consumption in the 12 months prior to the survey. Wave 1 contained only five of six items from the
USDA six-item food security scale collected each December in the Current Population Survey (CPS);
Wave 2 contained all six items from the December CPS module. Responses were used to assess
household food insecurity. Specifically, respondents were asked if “often/sometimes” in the last 12
months:
1. The food they bought just didn’t last, and they didn’t have money to get more.
2. They couldn’t afford to buy balanced meals.
3. They or other adults in household cut the size of their meals or skipped meals because there wasn’t enough money for food.
4. If yes to question 3, whether this happened almost every month, some months but not every month, or in only 1 or 2 months.
5. They ate less than they felt they should because there wasn’t enough money to buy food.
6. They were hungry but didn’t eat because they couldn’t afford enough food.
We sum the number of responses indicating “often” or “sometimes” to this battery of questions to
create a scale score reflecting household food insecurity. Households with scores of 0 or 1 are defined as
having high or marginal food security, households with scores of 2 to 4 indicate low food security, and
scores of 5 in Wave 1 or 5 to 6 in Wave 2 are defined as having very low food security. We define food
insecure households as those with either low or very low food security (summed scores ranging from 2 to
5 or 6, depending on wave). Households with summed scores of 0 or 1 are defined as food secure. We
classify respondents as experiencing persistent food insecurity if they report low or very low food security
in both waves of the MRRS. Ideally we would have the full six-item instrument in both waves, but
comparison of the five-item instrument to the full six items in Wave 2 indicates that the two instruments
provide nearly identical estimates of food insecurity (see Table 1). Unless specified, we use the five-item
measure in analyses reported below.
The MRRS collects self-reported information on receipt of benefits or assistance through SNAP,
TANF, Social Security Disability Insurance (SSDI), SSI, public health insurance such as Medicaid, UI,
14
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 1: Food Security among Households at or below 300% of the Federal Poverty Line (FPL) Percentage of Households with Income . . . Household Income and Composition in Wave 1 <= 100% of FPL 100–200% of FPL 200–300% of FPL Wave 1 Food Insecurity
0–1 Items 49.2ab 63.7a 66.6b 2–4 Items 30.6a 20.4a 24.8 5 Items 20.3b 15.8 8.6b Percent Food Insecure – 5-Item Measure 50.8ab 36.3a 33.4b
Wave 2 Food Insecurity 0–1 Items 56.3ab 70.1a 72.0b 2–4 Items 29.3 23.5 22.9 5 Items (Replicating Wave 1) 14.4ab 6.4a 5.2b 5–6 Items 19.0ab 9.0a 7.7b Percent Food Insecure – 5-Item Measure 43.7ab 29.9a 28.0b Percent Food Insecure – 6-Item Measure 43.7ab 30.3a 28.0b
Food Insecure – Wave 1 Only 17.7 16.6 12.6 Food Insecure – Wave 2 Only 10.6 10.2 7.2 Food Insecure Both Waves (5-Item Measure) 33.2a 19.7a 20.8 Notes: a,b,c – Cross-row cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession and Recovery Study (MRRS).
15
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
EITC, and public housing assistance. Also, the MRRS captures receipt of food or shelter assistance from
a private nonprofit charity in the prior year. Eligibility for SNAP is means-tested and assistance is
available only to households with income within 130 percent of the federal poverty line. Private charities
have greater discretion over eligibility, but often will target help to those with low-income. Accordingly,
we focus our analyses below on receipt of public safety net benefits and receipt of private support only for
respondents in households who reported annual income below 300 percent of the federal poverty line. As
household income rises above the poverty line, households are less likely to be eligible for public safety
net assistance programs, such as SNAP. Households with annual income near the poverty line, however,
may experience periods or spells during a year where income falls to the point where they may be income
eligible. Due to the effects of recession, these households coping with lower work earnings also may seek
help from nonprofit charities with less rigid income eligibility guidelines than public cash or in-kind
assistance programs.
Apart from income, race, age, presence of children, and marital status, our analyses focus on
several key household characteristics hypothesized to shape receipt of public and private supports. To
capture educational attainment, we categorize respondents by whether they have less than a high school
degree or equivalent, a high school degree or equivalent but no bachelor’s degree, or a completed
bachelor’s degree. Respondents indicated whether they were working or unemployed for each month of
the 12 months prior to the survey. We categorize respondents according to whether they were employed
in each of the previous 12 months, unemployed 1 to 6 months out of the previous 12, unemployed 7 or
more months, or whether they were out of the labor force for all twelve months prior to the interview. For
those out of the labor force in all 12 months, we distinguish between those with a physical health
limitation preventing them from working and those without such limitations.10 Finally, we use measures
of frequent church attendance, regular access to a car, and current or past union membership to reflect
10The MRRS includes a health limitation measure used in the Panel Study of Income Dynamics that reflects whether respondents “have any health problem or disability which prevents you from working or which limits the kind or amount of work you can do?”
16
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
community connectedness, and access to professional associations that may provide short-term help
during spells of unemployment.
We combine our survey data with local level data about the location of SNAP administrative
offices, food pantries, SNAP retailers, and food retailers that are used to create measures of food
assistance resource access. In doing so, we are in a unique position to connect household-level food
outcomes (program participation, food security, grocery shopping habits) to the local food resource
infrastructure with a precision not found in most published food policy research. Information about the
residential location of each MRRS respondent is used to locate households in a geographic information
system and street-grid, then to assess proximity and accessibility of different food assistance and retail
resources. Because few studies are able to make such textured spatial connections between households
and food resources, we examine several different measures of food resource access—each with its own
advantages and limitations. First, we examine distance to the nearest SNAP office, food pantry, SNAP
retailer, SNAP grocery store (as reported by the USDA), and grocery store or supermarket (as reported by
InfoUSA marketing data). While nearest-distance measures reflect immediate proximity, not all
households may shop or receive help at the closest store or provider. Distance to resources may operate as
a threshold effect, than linear decay process where greater distance steadily decreases probability of using
or visiting a particular resource. Thus, we create several distance bands to indicate whether a household
was located within 1, 2, or 3 miles of a given type of resource. Using data from the American Community
Survey (ACS), we examine the density of SNAP households within a mile of each respondent’s
household. We believe this local-level measure of food assistance may reflect levels of information about
food assistance and the degree of stigma associated with receiving assistance. Finally, nearest-distance
measures do not capture whether there are multiple resources nearby. The broader project, therefore,
considers the accessibility of food resources within different commuting times by different modes of
transportation. Here, we only examine the number of grocery stores or supermarkets (using InfoUSA
17
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
data) that fall within a 10-minute drive of each household.11 More specific information about the MRRS,
data sources used to calculate access, and the approach to calculating access can be found in the Technical
Appendix.
It is important to recognize that these survey data reflect a very unique urban setting. Whereas the
national unemployment rate hovered at 9 percent in August 2011, the unemployment rate for the Detroit
metropolitan area remained at nearly 12 percent—several percentage points above its pre-recession level.
And, at 36.2 percent in 2011, Detroit’s poverty rate was more than double the national average of 15
percent in 2011 (Bureau of Labor Statistics, 2013; DeNavas-Walt, Proctor, and Smith, 2012). The city
remains one of the most racially segregated urban centers in the country. Well-documented population
loss has left entire neighborhoods abandoned and led to a dramatic hollowing out of entire portions of the
central city. Compounding matters, the city of Detroit has experienced substantial fiscal problems over
the past several decades, culminating with its recent bankruptcy. Such fiscal problems make access to
certain types of locally financed public and private sources of support less predictable.
Alternatively, because the Detroit area has experienced difficult economic conditions longer than
most other metro areas, we might expect the lessons from Detroit may portend what can be expected in
other cities as their labor markets slowly recover. These data also will allow future work to use
geographic information about the location of respondents to accurately control for spatial access to a
number of different employment, safety net, and community resources when looking at household
outcomes.
Examining receipt of public and private sources of support in a single metropolitan area is
advantageous for several other reasons. Most importantly, by focusing on a single metro area in a single
state we are able to hold constant factors such as public program eligibility standards, benefit generosity,
and the strength of the local philanthropic community that vary from place to place. The SNAP program
in Michigan is called the “Food Assistance Program,” or FAP. Like other SNAP programs, FAP
11Roughly 90 percent of households in the MRRS report using a car when shopping for groceries.
18
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
eligibility is based on household composition, income, and an asset test added in 2011.12 Household
income must be less than 130 percent of the federal poverty line after a 20 percent earned income
disregard and deductions are calculated.13 The income limit is 200 percent of the federal poverty line if
the recipient is a senior or receives Social Security disability or Veteran’s Administration disability. A
household in which one member receives TANF assistance or State Disability Assistance (SDA) is
automatically eligible to receive FAP (Michigan Legal Help, 2013; State of Michigan, 2010). FAP
beneficiaries must recertify their benefits every 3, 12, or 24 months. Recertification requirements are
specific to each client, but generally require reporting of current income, assets, and expenses and
provision of verifying documents. Even though Detroit’s local labor market conditions may differ from
those of other areas, we believe the manner in which low-income households draw upon federal safety net
programs and patterns of use related to job loss and lost earnings should be generalizable to other settings.
To this point, the Detroit metropolitan region experienced increases in SNAP caseloads
commensurate with trends in other major urban areas and nationally. In 2000, there were 241,018 SNAP
recipients in Wayne County, compared to 30,139 in Oakland County and 19,090 in Macomb County.
SNAP receipt increased by 80 percent in Wayne County by 2009 (434,323 recipients), but increased by
213 percent in Oakland County (94,244 recipients) and by 402 percent in Macomb County (95,821
recipients). SNAP caseloads continued to increase through 2011, totaling more than 814,000 recipients in
the three-county metro area and nearly tripling between 2000 and 2011, according to USDA county-level
administrative data (U.S. Department of Agriculture, 2013a).14
12 In 2011, Michigan added asset tests to the eligibility criteria for FAP, which limited savings to $5,000 and value of a car to $15,000.
13 Income includes most kinds of earned and unearned income, such as wages, self-employment earnings, rental income, Social Security benefits, veterans’ benefits, and child support received. Apart from the standard deduction based on group size, household income calculations take into account shelter and dependent care deductions; see State of Michigan (2013b).
14In 2008, the average monthly FAP benefit per household was $212.38 and increased to $268.60 in 2010; see U.S. Department of Agriculture (2013a).
19
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
There is reason to expect that most MRRS respondents eligible for SNAP would have applied in
person and followed up on applications in person at one of 23 offices listed at the State of Michigan DHS
website in 2011. In recent years the State of Michigan has pursued SNAP modernization efforts that
include the creation of call centers, implementation of online eligibility screening, and completion of
applications online.15 Although Michigan has modernized its SNAP application process in order to
maximize outreach to potential SNAP households while minimizing administrative costs, these
modernization initiatives were not in place when the Waves 1 and 2 of the MRRS were in the field. For
example, in 2009 Michigan opened test call centers, but these only allowed current SNAP clients to report
changes (U.S. Department of Agriculture, 2010). In Michigan, online applications did not appear until
mid-2010.16 Finally, in 2010 the Michigan DHS began partnering with community organizations to test
self-service sign-up kiosks. Information kiosks and trained staff were placed in nonprofit social service
and food pantry locations, but the program did not become fully functional until 2011.17 Finally, SNAP
policy in Michigan still required applicants to appear for face-to-face interviews until 2011. Thus, current
and potential SNAP participants in 2008 and 2010 would still have been dependent on local SNAP
offices.
15Modernization efforts were in response to studies that showed that nonparticipating but eligible SNAP households cited the difficulty of getting to a food stamp office as a barrier to participation; see Bartlett and Burstein (2004).
16 Schwabish (2012) notes that although SNAP participation rates are higher in states with online applications, the impact of online applications on participation does not manifest itself immediately. Rather, participation grows minimally in the first three years after online applications are introduced, and accelerate thereafter.
17Michigan more intentionally engaged community-based organizations in outreach as part of its modernization, but a study by the USDA uncovered that applications submitted through community organizations constitute a “small minority of applications”; see U.S. Department of Agriculture (2013b).
20
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
EMPIRICAL RESULTS—RATES OF FOOD SECURITY AND PROGRAM PARTICIPATION IN METRO DETROIT
Food insecurity is quite prevalent among poor and near-poor households in metro Detroit
following the Great Recession.18 Nearly 51 percent of households below the federal poverty line were
food insecure in the year prior to the Wave 1 survey (50.8 percent, see Table 1). Twenty percent of poor
households—almost half of all poor households qualifying as food insecure—responded in the
affirmative to all five items, an indication of severe levels of food insecurity in the prior year. Consistent
with expectations that food insecurity is more common among nonpoor populations than is commonly
realized, 36.3 percent of households with income between 100 and 200 percent of the federal poverty line
and 33.4 percent of households with income between 200 and 300 percent of federal poverty were
classified as food insecure in the 12 months prior to Wave 1. Again, while these figures do not represent
food insecurity in all weeks or months of the preceding year, these high rates of food insecurity reflect the
economic and financial hardships that a wide range of poor and near-poor households grapple with in
metropolitan Detroit.
Possibly a reflection of the impact of economic recovery, the prevalence of food insecurity
declined slightly between waves of the MRRS across all income groups. The percentage of poor
households qualifying as food insecure (using the five-item measure) declined by about 14 percent
between waves (50.8 percent of households to 43.7 percent). Comparably sized declines in rates of food
insecurity occurred among near-poor households between the two waves. Across both waves of the
MRRS, however, food insecurity remained much more prevalent among poor households than near-poor
households.
18Overall, 23 percent of all MRRS respondents were food insecure at Wave 1 (Danziger, Allard, Wathen, Burgard, Seefeldt, Rodems, and Cohen, 2014). MRRS food insecurity estimates are comparable to estimates of population-level food insecurity generated by Gunderson, Waxman, Engelhard, Satoh, and Chawla (2013) in the three counties that compose the Detroit area in 2009: 23.8 percent for Wayne County; 15.3 percent for Oakland County; and, 17.7 percent for Macomb County.
21
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Finally, although a majority of food insecure households were food insecure in each wave of the
survey, we find some evidence of churning in and out of food insecurity in Detroit following the Great
Recession. For example, roughly one-third of those households reporting food insecurity in Wave 1 were
food secure in Wave 2. Similarly, about one-quarter of poor households that were food insecure in Wave
2, were food secure in Wave 1. One-third of poor households, however, were persistently food insecure
across the two waves of the survey, as were about 20 percent of households with income between 100 and
300 percent of poverty (see column 3). While the experience of food insecurity may be temporary for
some families, there is a sizeable percentage of poor and near-poor households who experience food
insecurity persistently over time.
As should be expected given income eligibility standards, receipt of SNAP and charitable food
assistance was more common among poor households than near-poor households in metro Detroit. In
large part this reflects income eligibility thresholds for SNAP (130 percent of federal poverty or 200
percent for seniors or those receiving Social Security disability or Veteran’s Administration disability).
Roughly 70 percent of poor households received SNAP in a given wave and more than 6 in 10 poor
households received SNAP in both waves (see top panel of Table 2, columns 1 through 3). Rates of
receipt are even higher among poor households with minor children, which is to be expected given that
the federal poverty line is determined by income and household size. About 85 percent of poor
households with children under 18 years of age reported receiving SNAP in the previous year.
Even households just above the poverty line or income eligibility thresholds may experience a
shock to household income at some point in the year that qualifies them to receive SNAP benefits.
Indeed, the second panel in Table 2 shows that 28.1 percent of households with income between 100 and
200 percent of federal poverty in Wave 1 reported receiving SNAP. As income moves above 200 percent
of poverty, receipt of SNAP is much less common with only 6.4 percent of households reporting such
assistance in Wave 1.
In contrast to the prevalence of food insecurity, we observe that the share of households receiving
SNAP increased slightly between waves—although the difference in SNAP receipt between waves does 22
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 2: Food Assistance Receipt among Households at or below 300% of the Federal Poverty Line (FPL) Percentage of Households Receiving SNAP Receiving Assistance from Charitable Nonprofits Household Income and Composition in Wave 1
Wave 1 Only (1)
Wave 2 Only (2)
Both Waves (3)
Wave 1 Only (4)
Wave 2 Only (5)
Both Waves (6)
<= 100% of FPL 6.1 8.8 a 61.7ab 12.5a 11.8a 21.1ab Households with Children under 18 3.3 6.1 81.4d 18.7b 16.3 21.2 Households with no Children under 18 8.2 10.9 46.3d 7.6b 8.2 21.0
100-200% of FPL 9.6 15.3a 18.5ac 7.9 8.4 5.7ac Households with Children under 18 14.2 14.5 14.3 6.5 12.5 6.6 Households with no Children under 18 5.5 16.0 22.3 9.1 4.8 4.8
200–300% of FPL 3.6 12.4 2.8bc 6.4a 4.6a 1.2bc Households with Children under 18 0.0 16.5 4.7 3.9 3.6 0.9 Households with no Children under 18 5.9 9.7 1.6 8.0 5.3 1.5
Notes: a,b,c – Within column cell-pair comparisons of FPL are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession and Recovery Study (MRRS).
23
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
not quite reach conventional levels of statistical significance. Nevertheless, these findings indicate that
receipt of SNAP food assistance in Detroit at least held steady almost two years after the recession
officially ended. Persistent receipt of food assistance is more common among households with income
below the federal poverty line. About 90 percent of poor households receiving SNAP in Wave 1 reported
SNAP benefits in both waves. We believe this reflects the low likelihood that poor households in Detroit
were able to find work or enough work to lift them above SNAP eligibility thresholds for an extended
period of time in the wake of the recession. As we also might expect, there is evidence that SNAP receipt
is more transitory among higher-income households. More frequent SNAP entry and exits likely reflect
short-term use by households that temporarily meet the income threshold in the course of the year.
Receipt of charitable food assistance is less common than SNAP assistance among poor and near-
poor households in Detroit. One-third of poor households report receiving charitable food assistance in
either wave of the MRRS (see top panel of Table 2, columns 4 through 6), compared to nearly 70 percent
of poor households who reported receiving SNAP in the previous year. Less than 15 percent of
households between 100 and 200 percent of poverty report receiving help from nonprofit charities in
Waves 1 or 2; less than 10 percent of households between 200 and 300 percent of poverty report help
from nonprofit charities in any given wave. Receipt of charitable food assistance also is less persistent
across the two waves of the MRRS compared to SNAP assistance. Approximately 2 of 3 poor households
reporting charitable food assistance received such assistance in both waves. Such findings fit with
evidence that nonprofit food pantries or soup kitchens often are a resource of last resort for households
experiencing persistent deep need or hardship (Feeding America, 2011).
Despite differences in prevalence and persistence, we find that the SNAP and charitable food
assistance caseloads in metropolitan Detroit share many similarities. Table 3 reports the demographic
characteristics of those on SNAP and receiving charitable food assistance. For instance, almost two-thirds
of SNAP recipients and recipients of nonprofit charity have income below the federal poverty line.
Reflecting the racial composition of Detroit and the disproportionately high rates of poverty among blacks
in Detroit, roughly 70 percent of SNAP and charity assistance recipients are black. More than three-24
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 3: Characteristics of MRRS Households at or below 300% of the Federal Poverty Level (FPL)
Household or Respondent Characteristic
All HHs <= 300% FPL
SNAP Recipients
Charity Recipients
Income at or below 100% of the FPL 35.1 61.9 63.3 Income 100% to 200% of the FPL 33.1 30.4 23.1 Income 200% to 300% of the FPL 31.8 7.7 13.6 Food Insecure 37.6 48.3 64.6 Food Secure 62.4 51.7 35.4 SNAP Recipients 38.1 — 67.8 No SNAP 61.9 — 32.2 Charity Recipients 17.7 31.8 — No Charity 82.3 68.2 — Black 43.7 68.5 73.5 Nonblack 56.3 31.5 26.5 HH with children 45.3 54.2 48.1 HH without children 54.7 45.8 51.9 Age 19–24 15.3 18.5 10.3 Age 25–34 21.9 23.4 21.0 Age 35–44 19.0 18.6 19.7 Age 45+ 43.8 39.5 49.0 Married 29.4 16.6 23.6 Not married 70.6 83.4 76.4 Less than HS 18.8 30.6 26.6 HS but no BA 65.7 63.6 67.5 BA or more 15.5 5.7 5.8 No unemployment 40.5 23.4 23.7 1–6 mos. unemployed 16.3 16.2 21.9 7–12 mos. unemployed 29.9 41.3 32.2 NILF all 12 months w/o health limitation 3.8 3.1 2.2 NILF all 12 months w/health limitation 9.5 15.9 20.0 Frequent Religious Attendance 41.1 40.4 41.5 Owns or Leases a Car 69.3 55.8 54.4 Was or is a Union Member 26.4 30.1 31.3 Note: Data are pooled across Waves 1 and 2 of the survey. Column percentages are reported. Household survey weights applied. Unweighted N = 969. Source: Michigan Recession & Recovery Study (MRRS).
25
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
quarters of assistance recipients are not married and more than 90 percent have a high school degree or
less.
Other interesting features of food assistance caseloads stand out in Table 3. First, 64.6 percent of
households receiving help from nonprofit charities also are food insecure, compared to 48.3 percent of
SNAP recipients. Again, this finding reflects the heightened financial distress associated with receipt of
food assistance from emergency service providers. Almost one-quarter of food assistance recipients—
both those reporting SNAP benefits or help from charitable nonprofits—reported no months of
unemployment in the prior year. Slightly more than half of SNAP recipients and households receiving
help from nonprofit charities experienced unemployment for more than 6 of the previous 12 months or
were out of the labor force altogether for the entire previous year. Finally, about 55 percent of food
assistance recipients have regular access to an automobile. Although not shown in Table 3, poor
households are much less likely to have regular access to a car than households with income above the
poverty line. Combined, these findings reflect the limited mobility of a large share of the food assistance
population in metropolitan Detroit, particularly those most in need or most at-risk of experiencing food
insecurity.
Food assistance programs may be important gateways to other types of safety net assistance.
Apart from categorical eligibility provisions in Michigan that automatically qualify those on TANF or
disability assistance for food stamps, the process of applying for SNAP may lead low-income households
to learn of other public programs for which they are eligible. Similarly, many food pantries or emergency
assistance providers offer clients information about public assistance programs and help with application
for public benefits (Allard, 2014). One possible pathway through which food assistance may reduce food
insecurity or economic hardship, therefore, may be through increasing take-up of other sources of
support. To begin to consider whether such pathways exist, Table 4 reports rates of program participation
for SNAP recipients and recipients of charitable food assistance across six types of public assistance:
TANF, SSI, public health insurance, EITC, UI, and public housing.
26
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 4: Bundling of Food Assistance among Households at or below 300% of the Federal Poverty Line (FPL) Percentage of Households that then Receive . . . Given that a household receives and has income: SNAP Charity TANF SSI
Public Health
Insurance EITC UI Public
Housing
Mean Number of
Public Programs
Public Program
and Charity
SNAP <100% of FPL — 36.9 30.0 41.1 67.2 55.1 9.2 30.3 2.9 36.9 100–200% of FPL — 25.6 12.9 29.7 50.9 53.7 21.3 14.6 2.5 25.6 200–300% of FPL — 15.1 26.5 19.7 19.5 35.8 25.6 13.4 2.2 15.1 Charity <100% of FPL 73.3 — 25.9 34.3 65.3 62.8 17.9 25.6 2.6 93.4 100–200% of FPL 68.4 — 8.1 34.3 50.2 41.4 14.5 9.7 2.0 83.4 200–300% of FPL 20.7 — 10.0 22.0 20.0 20.4 27.0 0.0 1.1 68.4 Note: Data are pooled across Waves 1 and 2 with household weights applied for SNAP and individual weights for Charity. Unweighted N = 969. Source: Michigan Recession & Recovery Study (MRRS).
27
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Although Table 4 reports simultaneous program participation, there is evidence that SNAP
recipients, poor and near-poor, draw upon additional public program resources. The average poor
household receiving SNAP also participates in almost two other public assistance programs. Even
nonpoor SNAP recipients receive help from slightly more than one other public program on average.
Sources of public support reflect components of the safety net that have become increasingly important to
working poor and near-poor families in the last decade. For example, nearly 70 percent of poor
households receiving SNAP also receive public health insurance coverage (67.2 percent) and slightly
more than half receive the EITC (55.1 percent). Thirty percent of SNAP recipients with income below the
poverty line also report receiving TANF at some point in the previous year. This is due to restrictive
TANF eligibility determinations and work requirements that lead many SNAP-eligible individuals to be
ineligible for TANF or to not apply altogether.
SNAP recipients with annual income within 200 percent of federal poverty also are likely to draw
upon public health insurance programs (50.9 percent) and the EITC (53.7 percent). About one in five
SNAP recipients with income between 100 and 200 percent of poverty also report household receipt of UI
in the previous year (21.3 percent), more than twice the rate observed among poor households receiving
SNAP. Table 4 also highlights how nonpoor households receiving SNAP draw upon other sources of
public assistance at some point during the year to cope with job loss or lost work earnings. Slightly more
than one-third of SNAP households with income between 200 and 300 percent of poverty report receiving
the EITC, one-quarter report receiving UI, and nearly 20 percent participated in a public health insurance
program.
Highlighting the important overlap of public and private sources of support within poor and near-
poor households, the average household reporting assistance from a charitable nonprofit participated in
more than two public assistance programs in the prior year. Roughly 70 percent of households at or below
200 percent of federal poverty reporting charitable food assistance also report SNAP benefits. These
figures reflect the depth of need among low-income households and gaps that public programs do not
cover. For example, SNAP clients may turn to charitable food pantries throughout the year to help stretch 28
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
food budgets or to provide meals after SNAP benefits for the month have been exhausted (Feeding
America, 2011). Two-thirds of poor households receiving help from nonprofit charities receive coverage
from public health insurance programs or assistance through the EITC.
Of significance for policy research concerned with spatial access to food assistance resources or
food retailers, Table 4 also underscores the extent to which food assistance populations are connected to
programs of disability assistance. More than 40 percent of poor households receiving SNAP also report
SSI benefits, as do nearly one-third of poor households receiving help from charitable nonprofits. In
Michigan, eligibility for SNAP is automatic for those receiving SSI when the recipient is an individual or
every member of his SNAP group is also receiving SSI. If some members of the group do not receive SSI,
categorical eligibility is forfeited, but SNAP eligibility is enhanced by a 200 percent federal poverty level
income limit, rather than the standard 130 percent limit (State of Michigan, 2013a).
Access to Food Resources
As we have shown, many food assistance households have limited access to transportation
resources. Such realities underscore the importance of living near food assistance programs and food
retailers. Tables 5 and 6 report mean food resource access score values across a number of different
measures by household income group, food security status, SNAP receipt, and receipt of charitable
assistance.
We find that poor households in Detroit live closer on average to SNAP administrative offices
and food pantries than households with income between 100 and 300 percent of poverty. For example,
households with income below the poverty line are about three-quarters of a mile closer to a SNAP
administrative office on average than households with income between 100 and 200 percent of poverty
(2.48 miles versus 3.22 miles, see column 1). Mean differences in the average distance to food pantries
follow a similar pattern, although the magnitude of the differences is much smaller. Poor households are
located about one mile from the closest food pantry on average, compared to 1.29 miles for households
just above the federal poverty line (see rows 1 and 2, column 2). Similarly, column 6 shows that poor 29
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
households live within one mile of about 500 more households receiving SNAP on average than
households with income between 100 and 200 percent of poverty (see column 6, 1,565 SNAP households
versus 1,011 SNAP households).
In contrast to expectations from the food desert literature, however, we find that poor households
on average are slightly closer to the nearest SNAP retailer, SNAP grocery store, or chain grocery store.
Poor households in metropolitan Detroit live about one-half mile from the closest SNAP-authorized
grocery store and about four-tenths of a mile from the closest chain supermarket or large grocery store on
average (see columns 4 and 5). By comparison, households with income between 200 and 300 percent of
poverty are located 0.8 miles from the nearest SNAP-authorized grocery store and 0.61 miles from the
closest supermarket or large grocery store on average (see columns 4 and 5). While statistically
significant, these differences in distance to the nearest grocery store amount to about two to three city
blocks.
The middle panel of Table 5 compares mean food resource access levels between poor
households that are food secure or insecure, as well as between poor households receiving or not
receiving food assistance. The bottom panel of Table 5 makes similar comparisons for households
between 100 and 200 percent of poverty. Several interesting findings emerge when we compare food
secure to insecure households or food assistance recipients to nonrecipients within the same income
strata. First, whether looking at poor households or near-poor households, food secure and food insecure
households do not differ significantly in most dimensions of food resource access. In fact, where mean
differences reach or approach conventional levels of statistical significance, we find that food secure
households are slightly farther from food assistance program resources and food retailers on average than
food insecure households. Such findings run counter to expectations that emerge from the food desert
literature, although the mean differences in access are quite small in size and may not reflect a meaningful
substantive difference.
We do find evidence that low-income households receiving SNAP are closer on average to both
SNAP administrative offices and food pantries. For example, SNAP recipients with income below the 30
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 5: Food Resource Access among Households at or below 300% of the Federal Poverty Line (FPL) by Food Security and Food Assistance in Wave 1 of the MRRS Measure of Food Resource Access in Wave 1 Average Distance in Miles to Nearest … Average Number of . . .
Wave 1 Household Characteristics
SNAP Office (1)
Food Pantry (2)
SNAP Retailer
(3)
SNAP Grocery
(4)
Grocery Store (5)
SNAP HHs within 1 mile
(6)
Grocery Stores within
10-minute drive (7)
Income <= 100% of FPL 2.48ab 0.99ab 0.28ab 0.52a 0.41a 1,565ab 1.23ab Income 100–200% of FPL 3.22a 1.29a 0.35ac 0.60b 0.52 1,011ac 0.90ac Income 200–300% of FPL 3.57b 1.45b 0.44bc 0.80ab 0.61a 744bc 0.63bc Income <= 100% of FPL
Food Insecure 2.29 1.06 0.27 0.53 0.40 1,582 1.34 Food Secure 2.67 0.92 0.28 0.51 0.41 1,548 1.13 SNAP Recipients 2.23a 0.86a 0.25a 0.48a 0.36a 1,714a 1.31 No SNAP 3.05a 1.29a 0.34a 0.61a 0.50a 1,199a 1.04 Charity Recipients 2.10a 0.77a 0.24 0.46 0.34a 1,859a 1.40a No Charity 2.67a 1.10a 0.30 0.55 0.44a 1,414a 1.15a
Income 100–200% of FPL Food Insecure 2.91 1.18 0.28a 0.47a 0.54 1,218a 0.99 Food Secure 3.40 1.36 0.39a 0.68a 0.50 889a 0.85 SNAP Recipients 2.55a 1.08 0.24a 0.45a 0.49 1,480a 1.29a No SNAP 3.49a 1.37 0.40a 0.66a 0.53 821a 0.75a Charity Recipients 2.64 1.10 0.28 0.51 0.53 1,469a 1.21 No Charity 3.31 1.32 0.36 0.62 0.52 937a 0.86
Notes: a,b,c,d,e,f – Cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession & Recovery Study (MRRS).
31
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
poverty line live almost one mile closer to a SNAP administrative office on average than poor households
not receiving SNAP (2.23 miles versus 3.05 miles, see column 1). Poor households receiving SNAP also
live about four-tenths of a mile closer to the nearest food pantry than poor households not receiving
SNAP (0.86 miles versus 1.29 miles, see column 2). Similarly, poor households receiving help from
nonprofit charities are located about one-half mile closer to SNAP administrative offices and about one-
third of a mile closer to the nearest food pantry than poor households not receiving help from such
charities.
Although the size of the mean differences is often quite small, we find evidence that poor
households receiving SNAP are located slightly closer to SNAP retailers and chain grocery stores than
poor households not receiving SNAP. As shown in column 4 of Table 5, poor households receiving
SNAP are 0.13 miles closer to a SNAP-authorized grocery store than a poor household not receiving
SNAP.
It may be that there are threshold distances or levels of access to food resources that matter more
than the nearest distance. For example, it may be more important to be located within a mile or two of a
food retailer or food assistance office, than to be within half a mile of a retailer or program of assistance.
Understanding the share of low-income households or households at-risk for food security located within
a few miles of food retailers or assistance programs may be more telling about disparities and gaps in
food resource access. Table 6 reports the share of households within certain distances of a SNAP
administrative office, grocery stores, and food pantries.
The top panel in Table 6 examines the share of households within certain distance bands of food
resources by income level. Consistent with Table 5, we find that poor households are disproportionately
likely to live within relatively short distances of SNAP offices and food pantries than nonpoor
households. For example, nearly half of all households in Detroit with income below the federal poverty
line are within 2 miles of a SNAP administrative office, compared to less than one-third of households
with income between 100 and 300 percent of federal poverty (column 2). While access may be greater
among poor households according to these measures, a large percentage of poor households remain more 32
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 6: Food Resource Access among Households at or below 300% of the Federal Poverty Line (FPL) by Food Security and Food Assistance in Wave 1 of the MRRS Percentage of Households in Wave 1 Distance to Nearest SNAP Office Within 1 Mile of a . . . Distance to Nearest Food Pantry Wave 1 Household Characteristics
Within 1 Mile (1)
Within 2 Miles
(2)
Within 3 Miles
(3)
SNAP Grocery
(4)
Grocery Store (5)
Within 1 Mile (6)
Within 2 Miles
(7)
Within 3 Miles
(8) Income <= 100% of FPL 14.4a 45.2a 68.8ab 87.8a 93.6 63.1ab 85.8 94.6 Income 100–200% of FPL 10.1b 29.1a 48.4a 86.9b 85.9 39.4a 82.1 93.2 Income 200–300% of FPL 3.5ab 26.9 45.7b 75.0ab 85.9 40.3b 73.2 95.0 Income <= 100% of FPL
Food Insecure 21.5a 54.9a 69.2 84.1 90.8 63.1 79.2 93.7 Food Secure 7.0a 35.1a 68.3 91.6 96.4 63.0 92.7 95.6 SNAP Recipients 16.5 49.0 76.8a 91.7a 97.2 70.3a 90.1 94.5 No SNAP 9.3 36.8 50.6a 78.5a 85.3 46.5a 75.6 94.7 Charity Recipients 20.6 56.5a 78.4a 90.2 95.8 69.4 91.7 100.0a No Charity 11.2 39.5a 63.9a 86.6 92.4 59.9 82.9 91.9a
Income 100–200% of FPL Food Insecure 15.2a 35.9 49.3 92.5 81.3 47.9 81.0 92.5 Food Secure 7.3a 25.2 47.9 83.7 88.5 34.6 82.7 93.6 SNAP Recipients 16.9a 46.9a 62.4a 92.3 87.9 61.8a 79.9 94.1 No SNAP 7.5a 22.1a 42.9a 84.8 85.1 30.7a 82.9 92.9 Charity Recipients 24.6a 45.2 56.2 91.5 81.3 58.4 74.7 100.0 No Charity 7.9a 26.6 47.2 86.2 86.6 36.4 83.2 92.2
Notes: a,b,c,d,e,f – Cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession & Recovery Study (MRRS)
33
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
than 3 miles from a SNAP office. Almost two-thirds of poor households are within one mile of a
nonprofit food pantry, compared to about 40 percent of households with income between 100 and 300
percent of federal poverty (column 6). When we expand the distance threshold a little farther, however,
we find that nearly every household with income at or below 300 percent of poverty is within 3 miles of a
food pantry in Detroit. At least when considering distances to nearest office locations, we find charitable
food assistance to be more accessible than SNAP offices.
SNAP benefits are placed on an electronic benefit transfer card (EBT) and used at food retailers.
It may be more important for SNAP recipients to be close to retailers and grocery stores that accept SNAP
EBT. Poor households appear to be slightly more likely to be located within one mile of a grocery or
supermarket that accepts SNAP than households between 200 and 300 percent of federal poverty (column
4). We also find that more than 8 in 10 households at or below 300 percent of poverty are within one mile
of grocery stores listed in the InfoUSA directory (column 5).
We observe similar patterns when examining the share of food insecure households or households
receiving food assistance that fall within a few miles of these same food resources. Food insecure
households with income below poverty are more likely to be within one or 2 miles of SNAP
administrative offices, compared to food secure households. Similarly, the bottom panel of Table 6 shows
that SNAP recipients and recipients of charitable food assistance are more likely to be within 3 miles of
SNAP administrative offices (column 3) and within one mile of nonprofit food pantries than nonrecipient
households in the same income bracket (column 6). SNAP recipients with household income below the
poverty line are slightly more likely to be within one mile of a grocery store accepting SNAP than poor
nonrecipients (91.7 percent versus 78.5 percent, respectively, column 4).
Tables 7 and 8 examine an identical set of food resource access measures, but across poor and
near-poor households according to whether they live in an urban or suburban location, have access to a
car, live in a tract where the poverty rate exceeds 20 percent, or live in a tract that is majority black. Of
significance, we find poor and near-poor suburban residents of metropolitan Detroit have far less access
to food assistance resources than comparably poor or near-poor residents of central city neighborhoods. 34
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
For example, poor households in the city of Detroit are 1.63 miles from a SNAP office and 0.57 miles
from a food pantry on average, compared to 3.39 miles to a SNAP office and 1.45 miles to a food pantry
for poor persons in suburban Detroit (see top panel of Table 7, columns 1 and 2). The top panel of Table 8
confirms these findings, indicating that less than half of poor respondents in suburban locations are within
3 miles of a SNAP office, compared to over 90 percent of poor respondents in the central city (44.1
percent versus 91.7 percent, respectively). Similarly, 90.3 percent of poor respondents in the central city
are within one mile of a food pantry, compared to 33.8 percent of poor respondents in the suburbs.
There is some evidence that poor and near-poor persons in suburban areas have less access to
food retailers than similar households in the central city, suggesting that food deserts may be more
problematic outside of cities than inside of cities. For example, poor persons in suburban areas are about a
quarter of a mile farther from a SNAP retailer on average than poor persons in the city of Detroit (0.40
miles versus 0.17 miles, respectively) and have access to about one-third as many grocery stores within a
10-minute drive (0.53 stores versus 1.88 stores, respectively).
Consistent with these findings, but in contrast to some of the research examining access to food
retailers, we find that residents in high-poverty and predominately black neighborhoods have both greater
access to food assistance resources and to food retailers. Poor and nonpoor residents of high-poverty
neighborhoods or predominately black neighborhoods are about twice as close to SNAP offices and food
pantries as residents of low-poverty or predominately non-black neighborhoods. Table 7 also shows that
poor and near-poor residents of high-poverty neighborhoods or predominately black neighborhoods are
slightly closer to SNAP retailers and have access to more grocery stores within a 10-minute drive than
residents of low-poverty or predominately non-black neighborhoods.
These initial examinations of food resource access provide some interesting observations about
the accessibility of food resources in the context of household income, food security, and food assistance.
First, food assistance resources appear to be located in closer proximity to poor households, food insecure
households, and to low-income households receiving food assistance, compared to their proximity to non-
poor households, food secure households, and those low-income households not receiving assistance. 35
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 7: Food Resource Access among Households at or below 300% of the Federal Poverty Line (FPL) by Neighborhood Characteristics and Regular Access to a Car in Wave 1 of the MRRS Measure of Food Resource Access in Wave 1 Average Distance in Miles to Nearest … Average Number of . . .
Wave 1 Household Characteristics
SNAP Office (1)
Food Pantry (2)
SNAP Retailer
(3)
SNAP Grocery
(4)
Grocery Store (5)
SNAP HHs within 1 mile
(6)
Grocery Stores within
10-minute drive (7)
Income <=100 of FPL Urban Resident 1.63a 0.57a 0.17a 0.51 0.33 2,254a 1.88a Suburban Resident 3.39a 1.45a 0.40a 0.54 0.49 801a 0.53a Household Has No Car 2.21a 0.80a 0.25 0.50 0.38 1,897a 1.46a Household Has Car 2.72a 1.17a 0.30 0.54 0.43 1,267a 1.02a High-Poverty Tract 1.74a 0.56a 0.16a 0.45 0.30a 2,136a 1.68a Low-Poverty Tract 3.52a 1.59a 0.45a 0.62 0.56a 729a 0.61a Majority Black Tract 1.97a 0.61a 0.20a 0.50a 0.34a 2,087a 1.60a Majority Non-black Tract 3.40a 1.70a 0.42a 0.57a 0.52a 563a 0.55a
Income 100–200% of FPL Urban Resident 1.58a 0.64a 0.18a 0.49 0.34a 2,317a 1.99a Suburban Resident 3.77a 1.51a 0.41a 0.64 0.58a 558a 0.54a Household Has No Car 2.82 1.31 0.32 0.56 0.48 1,285 1.12a Household Has Car 3.37 1.28 0.37 0.62 0.53 904 0.82a High-Poverty Tract 1.43a 0.69a 0.15a 0.43 0.28a 2,299a 1.97a Low-Poverty Tract 3.92a 1.52a 0.43a 0.67 0.61a 494a 0.49a Majority Black Tract 1.75a 0.71a 0.20a 0.51 0.35 2,237a 1.90a Majority Non-black Tract 3.78a 1.51a 0.41a 0.64 0.58 531a 0.53a
Notes: a,b,c,d,e,f – Cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession & Recovery Study (MRRS).
36
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 8: Food Resource Access among Households at or below 300% of the Federal Poverty Line (FPL) by Neighborhood Characteristics and Regular Access to a Car in Wave 1 of the MRRS Percentage of Households in Wave 1 Distance to Nearest SNAP Office Within 1 Mile of a . . . Distance to Nearest Food Pantry
Wave 1 Household Characteristics
Within 1 Mile (1)
Within 2 Miles
(2)
Within 3 Miles
(3)
SNAP Grocery
(4)
Grocery Store (5)
Within 1 Mile (6)
Within 2 Miles
(7)
Within 3 Miles
(8) Income <=100 of FPL
Urban Resident 18.3 71.6a 91.7a 92.6 100.0 90.3a 99.5a 100.0a Suburban Resident 10.1 16.7a 44.1a 82.7 86.7 33.8a 71.1a 88.9a Household Has No Car 15.0 51.6 77.3 92.2 97.7 74.5a 92.4 97.7 Household Has Car 13.8 39.2 60.8 83.7 89.7 52.4a 79.7 91.8 High-Poverty Tract 19.0 65.2a 93.9a 91.7 100.0a 91.9a 97.6a 98.1 Low-Poverty Tract 7.8 16.8a 33.3a 82.3 84.5a 22.4a 69.2a 89.8 Majority Black Tract 14.7 58.6a 87.2a 92.5 100.0 84.2a 97.4a 100.0a Majority Non-black Tract 13.8 20.4a 34.8a 79.1 81.8 24.1a 64.5a 84.8a
Income 100–200% of FPL Urban Resident 22.1 74.1a 95.6a 92.3 100.0a 84.4a 92.2 100.0 Suburban Resident 6.1 13.9a 32.4a 85.1 81.1a 24.3a 78.7 90.9 Household Has No Car 12.5 34.8 55.8 87.9 90.8 48.6 76.5 91.5 Household Has Car 9.2 26.9 45.6 86.5 84.0 35.9 84.2 93.9 High-Poverty Tract 29.5a 81.5a 97.3a 93.0 100.0a 84.5a 91.6 98.7 Low-Poverty Tract 2.6a 8.7a 29.4a 84.5 80.4a 21.9a 78.4 91.1 Majority Black Tract 20.2 70.8a 91.5a 92.9 100.0a 78.2a 91.0 100.0 Majority Non-black Tract 6.3 13.2a 32.0a 84.6 80.5a 24.7a 78.7 90.6
Notes: a,b,c,d,e,f – Cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the federal poverty line in both waves. Unweighted N = 485. Source: Michigan Recession & Recovery Study (MRRS).
37
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Second, poor and near-poor households receiving food assistance live amidst larger concentrations of
food assistance recipients than comparably poor households that do not receive assistance. While
disadvantaged households on average are slightly closer to food retailers of all types, most households are
within a mile of a grocery store or supermarket, whether they accept SNAP or not. Finally, inequality in
access to food resources appears to operate at the disadvantage of poor and near-poor residents of
suburban and low-poverty areas. Such findings run counter to previous research on food deserts and food
retailer access that use census tracts as the unit of analysis. Our snapshot here cannot uncover whether
food assistance recipients are more likely to locate near offices and relevant retailers, or whether offices
and retailers locate nearer food assistance recipients, but our findings suggest that low-income food
assistance recipients appear to have closer access to food resources than those low-income households
that do not receive assistance.
Factors Associated with Food Assistance Receipt
To better understand how demographic characteristics, economic shocks, and food resource
access is related to household receipt of food assistance, we estimated a set of probit models that examine
factors associated with receipt of SNAP or receipt of assistance from a nonprofit charity among
households with income at or below 300 percent of poverty. Results are reported in Tables 9 and 10.
As expected, several demographic characteristics are associated with household receipt of food
assistance. Black respondents were more likely to report receipt of SNAP and charitable assistance, as
were those respondents with low levels of educational attainment. We also find that households
experiencing any period of unemployment in the previous year are more likely to receive food assistance.
We also observe strong positive relationships between households not in the labor force where the
respondent has a health limitation and food assistance receipt.
We find access to food assistance resources and food retailers at times to be associated with
receipt of SNAP or charitable assistance. Table 10 reports probit coefficients for several different
measures of food resource access when these alternative access measures are included with the same set 38
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 9: Factors Associated with SNAP and Charitable Assistance Receipt among Households at or below 300% of the Federal Poverty Line (FPL), Probit Coefficients Reported
Receive SNAP
Receive Charitable Assistance
Distance to Closest SNAP Office -.014 (.035)
—
Distance to Closest Food Pantry —
-.045 (.077)
Respondent Race - Black .792** (.127)
.705** (.124)
Household with Children .578** (.170)
.205 (.134)
Respondent Age 19–24 Years Old .220
(.293) -.445* (.165)
25–34 Years Old .210 (.143)
-.162 (.178)
35–44 Years Old .159 (.207)
-.060 (.222)
Respondent Married -.541* (.196)
-.066 (.093)
Respondent Completed Education Less than High School 1.052**
(.219) .653*
(.257) High School but no B.A. degree .471**
(.167) .539**
(.207) Respondent Employment Status in Previous 12 Months
Unemployed 1–6 months .359** (.117)
.479+ (.254)
Unemployed 7–12 months .671** (.142)
.162 (.144)
NILF, no health limitation .218 (.269)
-.245 (.298)
NILF, with health limitation 1.347** (.273)
.708* (.297)
Frequent Religious Attendance .049 (.085)
-.021 (.131)
Owns or Leases a Car -.096 (.126)
-.162 (.119)
Was or is a Union Member -.153 (.128)
-.056 (.130)
Wave 1 -.264** (.092)
.100 (.108)
Constant -1.596** (.187)
-1.873** (.322)
N 949 960 Note: ** p< .01, * p< .05, + p< .10. Models were estimated using pooled Wave 1 and 2 data, household survey weights, and clustered standard errors. Reference categories for categorical predictors are: Age (45 and over), Education (B.A. or more), Unemployment (no unemployment). Standard errors are in parentheses. Source: Michigan Recession & Recovery Study (MRRS).
39
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 10: Relationship between Food Resource Access and Receipt of Assistance among Households at or below 300% of the Federal Poverty Line (FPL), Probit Coefficients Reported
Measure of Food Resource Access Receive SNAP
Receive Charitable Assistance
Distance to Closest SNAP Office -.014 (.035)
-.010 (.033)
Within 1 Mile of SNAP Office .362** (.126)
.340+ (.191)
Within 2 Miles of SNAP Office .121 (.145)
.069 (.150)
Within 3 Miles of SNAP Office .058 (.147)
.064 (.139)
Number of Households Receiving SNAP Within 1 Mile
.0002* (.0001)
.0001+ (.00006)
Distance to Closest SNAP Retailer -.381 (.270)
-.022 (.354)
Within 1 Mile of a SNAP Retailer -.070 (.267)
.285 (.440)
Distance to Closest SNAP Grocery Store or Supermarket
-.271* (.118)
-.014 (.153)
Within 1 Mile of SNAP Grocery Store or Supermarket .152 (.123)
-.080 (.208)
Distance to Closest Grocery Store or Supermarket -.219 (.132)
-.184 (.234)
Within 1 Mile of Grocery Store or Supermarket -.096 (.264)
.128 (.356)
Number of Grocery Stores or Supermarkets Within 10 Minute Drive
.110 (.108)
.004 (.055)
Distance to Closest Food Pantry -.080 (.138)
-.045 (.077)
Within 1 Mile of a Food Pantry .163 (.191)
-.074 (.135)
Within 2 Miles of a Food Pantry .064 (.288)
-.079 (.188)
Within 3 Miles of a Food Pantry -.172 (.412)
.742** (.223)
Note: ** p< .01, * p< .05, + p< .10. Models were estimated using pooled Wave 1 and 2 data, household survey weights, and clustered standard errors. Standard errors are in parentheses. Full model specification includes covariates reported in Table 9. Source: Michigan Recession & Recovery Study (MRRS).
40
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
of covariates reported in Table 9.19 While distance to the closest SNAP administrative office or food
pantry does not appear to be related to receipt of food assistance, we do find that living within one mile of
a SNAP office is positively related to both SNAP receipt and receipt of charitable assistance, even when
controlling for other household demographic and economic characteristics. Uncovering more evidence in
support of a threshold hypothesis for food resource access, we find that while living within a mile of a
food pantry has no significant association with receiving charitable food assistance, living within three
miles of a food pantry has a strong and positive relationship to receipt of assistance from nonprofit
charities.
To convey effect sizes, we translate our probit coefficients into predicted probabilities of
receiving SNAP or charitable assistance in Figures 1and 2 using a hypothetical baseline case and toggling
key right-hand side measures. Our baseline case is a not-married, nonblack household with children living
one half mile from a SNAP administrative office, between ages 25 and 34 with a high school degree at
Wave 1. The baseline household also has been employed in all twelve months prior to the survey, has a
car, frequently attends religious services, and is not a union member.
Predicted probabilities reported in Figure 1 highlight the importance of key household
characteristics in explaining food assistance receipt. Race and unemployment have large substantive
effects on food assistance receipt, nearly doubling the likelihood that a household receives SNAP and
increasing the likelihood of charitable assistance significantly. Consistent with the existing literature,
households where the respondent is out of the work force with a health limitation are more than twice as
likely to receive SNAP and more than three times as likely to receive charitable food assistance as the
baseline case.
19Coefficients for these other covariates are not reported in Table 10, although it should be noted that the substitution of different access measures into the baseline model presented in Table 9 does not significantly change the parameter estimates for other covariates.
41
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Figure 1: Predicted Probability of Receiving SNAP or Charitable Food Assistance, Demographic and Economic Characteristics
34.8%
65.6%
61.0%
83.0%
6.7%
21.4%
9.1%
21.5%
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Baseline HH Black Respondent Unemployed 7-12 mos. NILF w/health limitation
Pred
icte
d Pr
obab
ility
of R
ecei
ving
Ass
istan
ce
Household Type
SNAP
Charity
Source: Michigan Recession & Recovery Study (MRRS).
42
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Figure 2: Predicted Probability of Receiving SNAP or Charitable Food Assistance, Food Resource Access Measures
32.0%
45.8%
35.7%
43.8%
28.8%
34.2%
5.7%
10.7%
7.0%8.7%
1.2%
6.5%
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
More than 1 mile to SNAP office
Within 1 mile of SNAP office
1,000 SNAP HHs within 1 mile
2,000 SNAP HHs within 1 mile
> 1 mile from SNAP grocery store
< 1 mile from SNAP grocery store
> 3 miles from food pantry
< 3 miles of food pantry
Pred
icte
d Pr
obab
ility
of R
ecei
ving
Ass
ista
nce
Household Type
SNAP
Charity
Source: Michigan Recession & Recovery Study (MRRS).
43
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Figure 2 demonstrates the importance of spatial access to food resources.20 For example, being
within one mile of a SNAP administrative office increases the likelihood of receiving SNAP assistance by
about one-third over the baseline case where the household is more than a mile away from a SNAP office
(45.8 percent versus 32.0 percent). Closer proximity to a SNAP administrative office also is associated
with a higher likelihood of receiving charitable nonprofit food assistance (10.7 percent versus 5.7
percent).
NEXT STEPS AND IMPLICATIONS
Several important findings emerge from this project. First, we find food insecurity to be quite
prevalent among poor and near-poor households in metro Detroit following the Great Recession. Fifty-
one percent of households below the federal poverty line and 36 percent of households with income
below 200 percent of federal poverty were food insecure in the year prior to Wave 1 of the MRRS.
Consistent with effects of the economic recovery, the prevalence of food insecurity declined slightly
between waves of the MRRS. Nevertheless, about one-third of all households with income less than 300
percent of federal poverty were classified as food insecure in Wave 2.
Second, we find food assistance, SNAP in particular, to have become a critical source of support
for poor and near-poor households in metropolitan Detroit. Nearly 70 percent of households with income
below the poverty line report receiving SNAP benefits at some point in the prior year. About one-third of
households with income between 100 and 200 percent of federal poverty reported receiving SNAP in the
prior year. Receipt of charitable food assistance is less common among poor and near-poor households in
Detroit than is receipt of SNAP. About one-third of poor households report receiving charitable food
assistance and roughly 15 percent of households between 100 and 200 percent of poverty report receiving
help from nonprofit charities.
20Models are estimated separately for each measure of food resource access, so the predicted probability for the baseline case will be slightly across different measures of food resource access.
44
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Consistent with existing research, we find that being out of work and disconnected from the labor
market due to a physical health limitation are associated with increased household use of food assistance.
We find other household characteristics—race, presence of children, educational attainment, and marital
status—also to be associated with food assistance receipt.
Finally, this study finds evidence that low-income households receiving SNAP reside closer on
average to both SNAP administrative offices and food pantries than low-income households that do not
receive SNAP. There also is some evidence that poor households receiving SNAP are located slightly
closer to SNAP retailers and chain grocery stores than poor households not receiving SNAP. In contrast
to expectations from some of the existing literature on food deserts, however, we find that poor
households in Detroit on average are slightly closer to the nearest SNAP retailer, SNAP grocery store, or
chain grocery store than households in suburban areas.
We also find evidence that access to food assistance resources may be related to food assistance
receipt. Our results indicate that closer proximity or greater spatial access to food assistance resources
may be positively associated with SNAP take-up and food pantry receipt. Similarly, we find proximity to
larger numbers of SNAP recipients also increases the likelihood a household reports participation in
SNAP. Results presented here suggest that access to resources at or above a certain threshold affects
behavior or choice, suggesting that the relationship between food resource access and food outcomes may
be nonlinear.
Future analyses of these data will explore how household demographic characteristics, economic
shocks, food resource access, and food assistance are associated with grocery shopping habits and food
security in the MRRS. Tables 11 and 12 preview initial results from this ongoing research. Consistent
with other research, our research finds that nearly all households within 300 percent of federal poverty
report shopping at a grocery store or supermarket to purchase groceries (see Table 11). Households also
purchase food regularly from other types of food retailers, including specialty stores, convenience stores,
and drug stores. We find that about one-third of poor households switched the store where most grocery
shopping was done in the prior year. Many respondents cited prices, affordability considerations, 45
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 11: Grocery Shopping Behavior among Households at or below 300% of the Federal Poverty Line (FPL) in the MRRS Percentage of Households Where Household Buys Groceries
Switched Store or
Store Type
Why Switched Stores
Household Income and Access to Car in Wave 2
Grocery Store
Specialty Store
Convenience Store Drug Store
Prices/More Affordable/Can’t Afford Old Store More Convenient
Change in Transportation
<= 100% of FPL 96.9a 19.2 13.1 35.7a 32.5ab 60.1a 21.9 29.3a SNAP Recipients 94.7b 21.2 9.8 36.9 33.0 69.4 14.9 13.1b Charity Recipients 99.3b 17.0 16.7 34.5 32.0 49.8 29.6 47.2b
100–200% of FPL 99.2 27.6 8.5 23.0a 10.7ac 88.9a 44.6 17.6 SNAP Recipients 100.0 29.6 7.5 23.6 12.1 94.3 54.4a 18.0 Charity Recipients 97.1 22.2 11.4 21.4 7.0 64.4 0.0a 15.8
200–300% of FPL 100.0a 21.2 13.8 27.4 23.8bc 67.9 20.6 8.4a SNAP Recipients 100.0 22.5 14.4 28.5 26.7d 69.0b 20.9b 8.5c Charity Recipients 100.0 12.4 9.4 19.5 3.1d 0.0b 0.0b 0.0c
Notes: a,b,c – Within column cell-pair comparisons are statistically different at the .10 level or below. Household survey weights applied. Results reflect households that reported income at or below 300 percent of the national poverty line in both waves. Unweighted N = 485. Source: Michigan Recession and Recovery Study (MRRS).
46
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Table 12: Factors Associated with Food Insecurity among Households at or below 300% of the Federal Poverty Line (FPL)
Probit Bivariate
Probit IV Probit Receive SNAP .208
(.163) .091
(.502) 1.672
(1.117) Respondent Race - Black .189
(.115) .218
(.152) -.229 (.357)
Household with Children -.003 (.183)
.018 (.228)
-.267 (.256)
Respondent Age 19–24 Years Old -.511*
(.190) -.495* (.186)
-.507+ (.275)
25–34 Years Old .174 (.172)
.178 (.164)
.048 (.204)
35–44 Years Old .124 (.136)
.130 (.137)
.041 (.167)
Respondent Married -.096 (.210)
-.116 (.223)
.162 (.324)
Respondent Completed Education Less than High School 1.029**
(.201) 1.063** (.258)
.371 (.794)
High School but no B.A. degree .826** (.187)
.835** (.206)
.505 (.494)
Respondent Employment Status in Previous 12 Months
Unemployed 1–6 months .371+ (.205)
.382+ (.214)
.133 (.342)
Unemployed 7–12 months .274 (.226)
.294 (.264)
-.077 (.395)
NILF, no health limitation .397 (.540)
.398 (.540)
.244 (.568)
NILF, with health limitation .530+ (.272)
.573 (.361)
-.176 (.704)
Owns or Leases a Car .126 (.157)
.118 (.160)
.159 (.120)
Wave 1 .212* (.102)
.208+ (.107)
.261** (.081)
Constant -1.613** (.265)
-1.604** (.259)
-1.331* (.614)
Correlation of Error Terms – Bivariate Probit — .072 (.252)
—
N 949 945 949 Note: ** p< .01, * p< .05, + p< .10. Models were estimated using pooled Wave 1 and 2 data, household survey weights, and clustered standard errors. Reference categories for categorical predictors are: Age (45 and over), Education (B.A. or more), Unemployment (no unemployment). Standard errors are in parentheses. Source: Michigan Recession & Recovery Study (MRRS).
47
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
convenience, and changes in transportation as the main reasons behind these changes. Subsequent
analyses will examine how access to different types of food retailers is related to the mix of stores used
for grocery shopping and to shifts in this mix.
Additional research will examine whether SNAP receipt is associated with greater food security.
Again, we preview results from preliminary models in Table 12. Here we compare results from probit and
bivariate probit models predicting food security to a probit model using an instrumental variables
approach to address the selection into SNAP (see Daponte, Sanders, and Taylor, 1999; Gibson-Davis and
Foster, 2006; Gundersen and Oliveira, 2001; Nord and Golla, 2009). While each approach makes
different assumptions about the underlying relationship between food assistance and food security, receipt
of SNAP does not seem to be significantly associated with reported household food security.
Moving forward, we anticipate producing several manuscripts to submit to peer-reviewed
academic journals and reports targeted at the policy community. First, we will complete work on a
manuscript exploring the bundling of food assistance with other sources of public and charitable support.
Next, we will finish work on a manuscript that examines how household demographic characteristics,
economic conditions, and food resource access are associated with receipt of SNAP and food pantry
assistance. Finally, we will produce two additional manuscripts that examine the determinants of food
shopping behaviors and food security more directly. As we complete these manuscripts, we hope to
submit discussion papers to the Institute for Research on Poverty (IRP) and to the National Poverty
Center, as well as briefs for the Focus and Fast Focus series at IRP.
48
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
TECHNICAL APPENDIX
Michigan Recession and Recovery Survey (MRRS). Determinations of household low-income
status in Wave 1 are based on income for calendar year 2008 and for Wave 2, calendar year 2010.
Household income is drawn from a single question and verified through checking responses to different
sources of income. In the single question, MRRS asked each respondent, “Now thinking about you and
your household, what do you estimate was the total income in 2008 for you and all other people living
here from all sources, including earnings from work, any business, plus food stamp payments, child
support, any government benefits, retirement income and any interest or investment income, before
taxes?” For Wave 1, the federal poverty threshold for a single-parent household with two children in 2008
was $17,600. Households within 300 percent of this threshold have annual income at or below $52,800.
Pooled across the two waves, 42.9 percent of MRRS respondents reported household income within 300
percent of the federal poverty line. Table 2 reports the characteristics of MRRS respondents within 300
percent of the federal poverty line using data pooled across the two waves of the survey.
Respondents indicated whether they were receiving public health insurance and housing
assistance at the time of the survey. Measures of TANF, SNAP, UI, and EITC receipt are based on
whether respondents received such assistance in the past 12 months. The MRRS does not allow us to
assess in which months households are eligible for programs and in which months they receive program
assistance. Rates of public program receipt have been benchmarked against 2009 Current Population
Survey data reported in the March 2010 Survey for the three-county Metropolitan Detroit area. We find
reported rates of program participation to be very similar between the MRRS and CPS. In addition, all
measures of public benefit receipt are weighted using household survey weights, with the exception of
public health insurance receipt. Public health insurance was asked of individual respondents rather than
whether anyone in the household had public health insurance, so we use individual survey weights when
reporting rates of public health coverage.
49
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Food Pantry Survey. A list of 407 charitable nonprofit food pantries or emergency food
programs located in the study area of the MRRS were compiled from online directory listings and the
United Way of Southeastern Michigan 2-1-1 directory in Spring 2012. A letter of invitation to participate
in a short survey was sent to each listed pantry. A 10-minute telephone survey instrument collecting
information about location, program services, client characteristics, and funding was developed and pilot-
tested with five Chicago-area food pantries. Survey call attempts began in August 2012 and were
completed in April 2013. Surveys were completed at the Population Research Center at NORC and the
University of Chicago by a trained telephone survey interviewer.
When reaching a food assistance program, the survey interviewer asked to speak to the program
executive or to a program manager that could answer some basic questions about the programs available
on-site. Many organizations were not eligible for the survey: 37 were no longer operational; 29 were not
food assistance programs; contact information could not be located for nine other listings. Surveys were
completed with 263 of the remaining 332 listed programs for a response rate of 80.2 percent. Twelve
programs refused to participate in the survey and 57 programs were never reached to complete calls. All
organizations not completing surveys were contacted at least 10 times by the interviewer, but only 37 of
the 57 programs not reached appeared to have a functioning phone system. A total of 1,674 call attempts
were made.
Providers offered a range of services to low-income individuals on-site. Nearly 90 percent offered
groceries—most through a food pantry progam. One-third provided meals on-site to low-income and
three-quarters provided non-food related benefits (e.g., housing assistance or shelter, utility assistance,
clothing or furniture). Some organizations provided job training, health services, and referrals to other
social service providers. Nearly half—49 percent—reported helping clients connect to public assistance
programs for which they may be eligible. The average food assistance program served 1,134 individuals
in a typical month, although the median provider served 400 in a typical month. Programs averaged 1.5
FTE staff and about 100 volunteer hours per week.
50
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
InfoUSA Food Retailer Data. Lists of food retailers operating in the Detroit metropolitan area
were obtained from InfoGroup for 2008 and 2010. Files for both years contain information on businesses
operating at the end of each year. The business selection criteria were as follows: (1) geographic area:
Wayne, Oakland, and Macomb counties in Detroit Metropolitan Area (Michigan) PLUS the adjacent 32
zip code areas; and (2) industrial classification codes (NAICS): 44511, 44512, and 4452. In 2008, there
were 2,818 food retail businesses fitting our search criteria and 2,860 food retail businesses matching
search criteria in 2010.
SNAP Administrative Office Locations. Office locations where applications for SNAP may be
submitted and processed in the three-county Detroit metropolitan area. Information about office locations
(N = 23) were drawn in March 2011 from the State of Michigan Department of Human Services (DHS)
website (http://www.michigan.gov/dhs/). Geographic coordinates of each office were then added by
geocoding the locations in ArcGIS.
SNAP Retailer Data. Lists of authorized SNAP retailers in the State of Michigan for the years
2008 and 2010 were obtained from the USDA’s Food and Nutrition Service program
(http://www.fns.usda.gov/fns/) via email on October 15, 2012. These lists represent retailers in Michigan
that are in an authorized status to receive SNAP at the end of the Fiscal Year (09/30/08 for 2008 data and
9/30/10 for 2010 data). A team of research assistants from the University of Michigan examined each
SNAP retailer, entered the address into Google Maps, and located the building in street view. Using the
image of the store and the store name, the team coded each retailer into one of seven categories: Grocery
store/Chain Grocery (i.e., Kroger); Drug Store/Dollar Store/Chain Retail (i.e., Walgreens, CVS, Target,
Dollar Store, Kmart); Gas Station; Mini-mart/Convenience Store/Liquor/Party Store;
Bakery/Butcher/Other Specialty Foods; Farmers Market; Other. Coding of stores was check-coded for
consistency. Store addresses were geocoded and geographic coordinates of each store were then added to
the GIS.
Food Resource Access Measures. Food resource access measures were calculated by
determining time and distance between each MRRS respondents’ home and a given food resource. Food 51
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
resources include location of SNAP administrative offices, food pantries, SNAP retailers, and food
retailers regardless of their authorization to receive SNAP. Addresses of the respondents from two waves
are geocoded to 2010 Census geography. To calculate access measures, the sum of the number of food
retailers or food assistance programs of a given type located within a reasonable commute time (5-, 10-,
15-, 20-, 30-minute) of each MRRS respondent’s home was calculated for three different travel modes:
driving; public transit; and walking. In addition, Euclidean or straight-line distance in miles to the nearest
food resource of a given type for a given mode also was calculated.
The least cost driving time was calculated in ArcGIS using the routing service data from the
StreetMap North America dataset 10.1, and under the following assumptions. To calculate the driving
time, we restricted the use of private entryways, waterways, and toll roads and asked the software to
choose a route to minimize travel time (i.e., time as the impedance), or the quickest route. Walking time
was also calculated in ArcGIS and using the same routing service data, and under almost the same
assumptions, except one, which restricts the use of limited access roads, namely highways, as it is
assumed that pedestrians cannot walk along such roads. The software chose the route that minimized
travel distance, or that represented the shortest route. Once the shortest travel route was assigned for each
trip between the participant’s residence and all destinations, the distance was then translated into travel
time, assuming an average comfortable walking speed of 3 miles per hour. Calculation of the public
transit travel time was done in Stata, where a “TRAVELTIME” command file retrieves estimated public
transit travel time using the Google Transit web service and assuming the Detroit public transportation
system (SMART) as the mode of transportation. It should be noted that public transit travel time varies by
time and day of the travel. The time of the travel we used was 9:00 a.m., and the day of the travel was
Wednesday, September 26, 2012.
52
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
REFERENCES
Adams, Terry K., Jim Lepkowski, Mahmoud Elkasabi, and Danielle Battle. (2011). “Michigan Recession and Recovery Study (MRRS): Sampling and Weights Documentation.” University of Michigan, Institute for Social Research.
Algert, Susan J., Michael Reibel, and Marian J. Renvall. 2006. “Barriers to Participation in the Food Stamp Program among Food Pantry Clients in Los Angeles.” American Journal of Public Health, 96(5): 807–09.
Allard, Scott W. 2008. “Place, Race, and Access to the Safety Net.” In Colors of Poverty, eds. Ann Chih Lin and David Harris. New York: Russell Sage Foundation.
Allard, Scott. W. 2009a. “Mismatches and Unmet Needs: Access to Social Services in Urban and Rural America.” in Welfare Reform and its Long-Term Consequence for America’s Poor, James P. Ziliak (ed.), Cambridge, UK: Cambridge University Press.
Allard, Scott W. 2009b. Out of Reach: Place, Poverty, and the New American Welfare State. New Haven: Yale University Press.
Allard, Scott. W. 2014. “Placing Food Security in a Spatial Context.” Institute for Research on Poverty Working Paper.
Allard, Scott W. and Benjamin Roth. 2010. “Strained Suburbs: The Social Service Challenges of Rising Suburban Poverty.” Washington, D.C.: The Brookings Institution Metropolitan Policy Program.
Allard, Scott W. and Benjamin Roth. 2014. “The Uneven Geography of Latino Immigrant Settlement and Social Services.” Working paper.
Allard, Scott W., Richard M. Tolman and Daniel Rosen. 2003. “Proximity to Service Providers and Service Utilization among Welfare Recipients: The Interaction of Place and Race.” Journal of Policy Analysis and Management 22(4): 599–613.
Bartfeld, Judi. 2003. “Single Mothers, Emergency Food Assistance, and Food Stamps in the Welfare Reform Age.” The Journal of Consumer Affairs 37 (2): 283–303.
Bartfeld, Judi and Rachel Dunifon. 2006. “State-Level Predictors of Food Insecurity among Households with Children.” Journal of Policy Analysis and Management 25(4): 921–42.
Bartlett, Susan and Nancy Burstein. 2004. Food Stamp Program Access Study: Eligible Nonparticipants, report prepared by Abt Associates Inc. under a research contract from the Economic Research Service (May).
Bartlett, Susan, Nancy Burstein, and William Hamilton. 2004. “Food Stamp Program Access Study: Final Report.” Economic Research Service, Report #E-FAN-03-013-3.
Bhattarai, Gandhi Raj, Patricia Duffy, and Jennie Raymond. 2005. “Use of Food Pantries and Food Stamps in Low-Income Households in the United States.” The Journal of Consumer Affairs, 39(2):276–98.
53
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Borjas, George J. 2004. “Food insecurity and Public Assistance.” Journal of Public Economics, 88(7–8):1421–43
Bureau of Labor Statistics. (2013). “Smoothed Seasonally Adjusted Metropolitan Area Estimates.”
Center on Budget and Policy Priorities. 2010. “Policy Basics: Introduction to the Food Stamp Program.” Accessed March 27, 2011. http://www.cbpp.org/files/policybasics-foodstamps.pdf.
Center on Budget and Policy Priorities. 2012. “Policy Basics: Introduction to the Food Stamp Program.”
Center on Budget and Policy Priorities. 2013. “Policy Basics: Introduction to the Supplemental Nutrition Assistance Program (SNAP).” Accessed November 2, 2013. http://www.cbpp.org/files/policybasics-foodstamps.pdf
Chaudry, Ajay and Karina Fortuny. 2010. Children of Immigrants: Economic Well-being (Brief No. 4). Washington, D.C.: Urban Institute.
Coleman-Jensen, Alisha, Mark Nord, Margaret Andrews, and Steven Carlson. 2012. Household Food Security in the United States in 2011. U.S. Department of Agriculture, Economic Research Service, ERR-141.
Congressional Budget Office (CBO). 2007. “Supplemental Security Income, CBO January 2007 Baseline.”
Congressional Budget Office (CBO). 2012. “Supplemental Security Income, CBO January 2012 Baseline.”
Danziger, Sandra K., Scott W. Allard, Maria V. Wathen, Sarah A. Burgard, Kristin S. Seefeldt, Rick Rodems, and Alicia Cohen. 2014. “Food Insecurity in the Detroit Metropolitan Area Following the Great Recession.” University of Michigan National Poverty Center, Policy Brief.
Daponte, Beth Osborne, Seth Sanders, Lowell Taylor. 1999. “Why Do Low-Income Households not Use Food Stamps? Evidence from an Experiment.” The Journal of Human Resources 34(3): 612–28.
DeNavas-Walt, Carmen, Bernadette D. Proctor, Jessica C. Smith. 2011. Income, Poverty, and Health Insurance Coverage in the United States: 2010. Census Bureau, Current Population Reports #P60-239;
DeNavas-Walt, Carmen, Bernadette D. Proctor, Jessica C. Smith. 2012. Income, Poverty, and Health Insurance Coverage in the United States: 2011. Census Bureau, Current Population Reports #P60-243.
Eslami, Esa, Kai Filion, and Mark Strayer. 2011. “Characteristics of Supplemental Nutrition Assistance Program Households: Fiscal Year 2010.” United States Department of Agriculture, Food and Nutrition Service, Supplemental Nutrition Assistance Program, Report No. SNAP-11-CHAR.
Feeding America. 2011. “Food Banks: Hunger’s New Staple.”
54
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Fix, Michael E., Randy Capps, and Neeraj Kaushal. 2009. “Immigrants and Welfare: Overview.” In Immigrants and Welfare: The Impact of Welfare Reform on America’s Newcomers, Michael E. Fix (ed.), New York: Russell Sage Foundation Press.
Gallagher, Mari. 2006. “Examining the Impact of Food Deserts on Public Health in Chicago.” Chicago: Mari Gallagher Research & Consulting Group. Accessed March 27, 2011. http://www.marigallagher.com/site_media/dynamic/project_files/1_ChicagoFoodDesertReport-Full_.pdf.
Garr, Emily. 2011. “The Landscape of Recession: Unemployment and Safety Net Services across Urban and Suburban America.” Washington, D.C.: The Brookings Institution Metropolitan Policy Program.
Gibson-Davis, Christina M. and E. Michael Foster. 2006. “A Cautionary Tale: Using Propensity Scores to Estimate the Effect of Food Stamps on Food Insecurity.” Social Service Review 80(2006): 93–126.
Gundersen, Craig and Victor Oliveira. 2001. “The Food Stamp Program and Food Insufficiency.” American Journal of Agricultural Economics 83(4): 875–87.
Gunderson, Craig, Elaine Waxman, Emily Engelhard, Amy Satoh and Namrita Chawla. Map the Meal Gap 2013. Technical Brief. http://feedingamerica.org/hunger-in-america/hunger-studies/map-the-meal-gap.aspx
Haider, Steven J. and Marianne Bitler. 2009. “An Economic View of Food Deserts in the United States.” Paper presented at the National Poverty Center conference on Understanding the Economic Concepts and Characteristics of Food Access, January 2009.
Hanratty, Maria J. 2006. “Has the Food Stamp Program Become More Accessible? Impacts of Recent Changes in Reporting Requirements and Asset Eligibility Limits.” Journal of Policy Analysis and Management, Vol. 25, No. 3, 603–21.
Isaacs, Julia B., Tracy Vericker, Jennifer Macomber, and Adam Kent. 2009. “Kids’ Share: An Analysis of Federal Expenditures through 2008.” Washington, D.C.: Urban Institute and Brookings Institution.
Johnson, Janna. 2012. “Supplemental Nutrition Assistance Program Participation during the Economic Recovery of 2003 to 2007.” Focus, 29(1): 9–13.
Kaiser Commission of Medicaid and the Uninsured. 2012. “Medicaid Enrollment: June 2011 Data Snapshot.”
Keane, Michael and Robert Moffitt. 1998. “A Structural Model of Multiple Welfare Program Participation and Labor Supply.” International Economic Review, 39(3): 553–89.
Kim, J., Shelley K. Irving, and Tracy A. Loveless. 2012. “Dynamics of Economic Well-Being: Participation in Government Programs, 2004 to 2007 to 2009—Who Gets Assistance?” U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau, Current Population Reports #P70-130.
55
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Kissane, Rebecca Joyce. 2003. “What’s Need Got to Do with It? Barriers to Use of Nonprofit Social Services.” Journal of Sociology and Social Welfare 30(2):127–48.
Klerman, Jacob A. and Caroline Danielson. 2011. “The Transformation of the Supplemental Nutrition Assistance Program.” Journal of Policy Analysis and Management, Vol. 30, No. 4, 863–88.
Klerman, Jacob A. and Caroline Danielson. 2012. “What Caused the Rise in SNAP Participation?” Abt Associates, Inc.
Kneebone, Elizabeth. 2009. “Economic Recovery and the EITC: Expanding the Earned Income Tax Credit to Benefit Families and Places.” Washington, D.C.: The Brookings Institution Metropolitan Policy Program.
Martin, Katie S., John T. Cook, Beatrice L. Rogers, and Hugh M. Joseph. 2003. “Public versus Private Food Assistance: Barriers to Participation Differ by Age and Ethnicity.” Journal of Nutrition Education and Behavior, 35(5):249–54.
Mathematica Policy Research Inc. 2010. “Hunger in America 2010 National Report.” Accessed March 27, 2011. http://feedingamerica.issuelab.org/research/listing/hunger_in_america_2010_national_report.
McKernan, Signe-Mary and Caroline Ratcliffe. 2003. “Employment Factors Influencing Food Stamp Program Participation.” The Urban Institute.
Michigan Legal Help. Food Assistance Program (FAP, or Food Stamps). Retrieved November 11, 2013, from http://michiganlegalhelp.org/self-help-tools/public-benefits/food-assistance-program-fap-or-food-stamps.
Moffitt, Robert A. 2012. “The Social Safety Net and the Great Recession.” Russell Sage Foundation and the Stanford Center on Poverty and Inequality.
Moore, Latetia and Ana V. Diez Roux. 2006. “Associations of Neighborhood Characteristics with the Location and Type of Food Stores.” American Journal of Public Health 96(2): 1–7.
Neckerman, Kathryn M., Michael Bader, Marnie Purciel, Paulette Yousefzadeh. 2009. “Measuring Food Access in Urban Areas.” Paper presented at the National Poverty Center conference on Understanding the Economic Concepts and Characteristics of Food Access, January 2009.
Nord, Mark, Margaret Andrews, and Steven Carlson. 2008. “Household Food Security in the United States, 2007.” U. S. Department of Agriculture, Economic Research Service. Economic Research Report No. 66.
Nord, Mark and Anne Marie Golla. 2009. “Does SNAP Decrease Food Insecurity?” U. S. Department of Agriculture, Economic Research Service. Economic Research Report No. 85.
Pan, Suwen and Helen Jensen. 2008. “Does the Food Stamp Program Affect Food Security Status and the Composition of Food Expenditures?” Journal of Agricultural and Applied Economics, 40(1): 21–35.
56
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Powell, Lisa M., Sandy Slater, Donka Mirtcheva, Yanjun Bao, Frank J. Chaloupka. 2007. “Food Store Availability and Neighborhood Characteristics in the United States.” Preventive Medicine 44 (2007): 189–95.
Purtell, Kelly M., Elizabeth T. Gershoff, J. Lawrence Aber. 2012. “Low Income Families’ Utilization of the Federal ‘Safety Net’: Individual and State‐Level Predictors of TANF and Food Stamp Receipt.” National Poverty Center Working Paper Series, #12-04.
Ratcliffe, Caroline, Signe-Mary McKernan, and Kennth Finegold. 2008. “Effects of Food Stamp and TANF Policies on Food Stamp Receipt.” Social Service Review, 82(2): 292–334;
Reese, Kanin L. 2006. “An Analysis of the Characteristics of Multiple Program Participation Using the Survey of Income and Program Participation (SIPP).” U.S. Department of Commerce, U.S. Census Bureau, SIPP Working Paper No. 244.
Rosenbaum, Dorothy. 2013. “The Relationship Between SNAP and Work Among Low-Income Households.” Center on Budget and Policy Priorities.
Schmidt, Lucie. 2012. “The Supplemental Security Income Program and Welfare Reform.” Federal Reserve of Boston, Public Policy Discussion Paper No. 12-3.
Schwabish, Jonathan. Downloading Benefits: The Impact of Online Food Stamp Applications on Participation. Paper presented at the Association for Public Policy Analysis and Management conference, November 10, 2012.
Shaefer, H. Luke, and Italo Gutierrez. 2012. “The Supplemental Nutrition Assistance Program and Material Hardships among Low-Income Households with Children.” National Poverty Center Working Paper, #11-18.
Simms, Margaret C. 2008. “Weathering Job Loss: Unemployment Insurance.” Washington, D.C.: Urban Institute; Tax Policy Center, Urban Institute, and Brookings Institution. “Spending on the EITC, Child Tax Credit, and AFDC/TANF, 1976–2010.”
Smith, Vernon K., Kathleen Gifford, Eileen Ellis, Robin Rudowitz and Laura Snyder. 2012. “Medicaid Today; Preparing for Tomorrow A Look at State Medicaid Program Spending, Enrollment and Policy Trends.” Kaiser Commission of Medicaid and the Uninsured.
Soss, Joe, Richard C. Fording, and Sandford F. Schram (2011). Disciplining the Poor. University of Chicago Press.
Sparks, Andrea L., Neil Bania and Laura Leete. 2011. “Comparative Approaches to Measuring Food Access in Urban Areas: The Case of Portland, Oregon.” Urban Studies, 48(8): 1715–1737.
State of Michigan, Department of Human Services. 2010. Michigan Combined Application Project. (DHS-PUB-0500-271413-7). Retrieved November 11, 2013, from www.michigan.gov/documents/dhs/DHS-PUB-0500_271413_7.pdf.
State of Michigan, Department of Human Services. 2013a. Bridges Eligibility Manual: Categorical Eligibility. (BPB 2013-011) Accessed January 7, 2014. http://www.mfia.state.mi.us/olmweb/ex/html/
57
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
State of Michigan, Department of Human Services. 2013b. FAP Income Limits. (RFT 250/RFB 2013-007). Retrieved November 11, 2013, http://www.mfia.state.mi.us/olmweb/ex/rft/250.pdf.
US Department of Health and Human Services. 2010. “TANF Financial Data.” U.S. House of Representatives, House Committee on Ways and Means. 2004. 2004 Green Book.
US Conference of Mayors. 2008. “Hunger and Homelessness Survey: in America’s Cities: A Status Report on Hunger and Homelessness in America’s Cities.”
US Department of Health and Human Services. 2010. “TANF Financial Data.” U.S. House of Representatives, House Committee on Ways and Means. 2004. 2004 Green Book.
US Conference of Mayors. 2008. “Hunger and Homelessness Survey: in America’s Cities: A Status Report on Hunger and Homelessness in America’s Cities.”
US Department of Agriculture. 2009a. “Access to Affordable and Nutritious Food: Measuring and Understanding Food Deserts and Their Consequences.” Accessed June 18, 2013. http://www.ers.usda.gov/media/242675/ap036_1_.pdf.
US Department of Agriculture, Food and Nutrition Service, Office of Research and Analysis. 2009b. “Understanding the Determinants of Supplemental Nutrition Assistance Program Participation,” by Nancy R. Burstein, Satyendra Patrabansh, William L. Hamilton, and Sarah Y. Siegel. Project Officer: Rosemarie Downer, Alexandria, VA.
US Department of Agriculture, Food and Nutrition Service, Office of Research and Analysis. 2010. “Enhancing Supplemental Nutrition Assistance Program (SNAP) Certification: SNAP Modernization Efforts: Interim Report - Volume 1,” by Gretchen Rowe, Sam Hall, Carolyn O’Brien, Nancy Pindus, and Robin Koralek. Project officer: Rosemarie Downer, Alexandria, VA.
US Department of Agriculture, Food and Nutrition Service, Office of Research and Analysis. 2012a. “Characteristics of Supplemental Nutrition Assistance Program Households: Fiscal Year 2011,” by Mark Strayer, Esa Eslami, and Joshua Leftin. Project Officer, Jenny Genser. Alexandria, VA.
US Department of Agriculture, Food and Nutrition Service, Office of Research and Analysis. 2012b. “Building a Healthy America: A Profile of the Supplemental Nutrition Assistance Program.”
US Department of Agriculture, Economic Research Service. 2013a. “Supplemental Nutrition Assistance Program (SNAP) Data System.” http://www.ers.usda.gov/data-products/supplemental-nutrition-assistance-program-(snap)-data-system/time-series-data.aspx#.UnAgtnCsiCc. Last Accessed October 26, 2013.
US Department of Agriculture, Food and Nutrition Service, Office of Research and Analysis. 2013b. “The Evolution of SNAP Modernization Initiatives in Five States,” by Lara Hulsey, Kevin Conway, Andrew Gothro, Rebecca Kleinman, Megan Reilly, Scott Cody, and Emily Sama-Miller. Project Officer, Rosemarie Downer. Alexandria, VA.
US Department of Agriculture, Food and Nutrition Service. 2013c. Supplemental Nutrition Assistance Program: Average Monthly Benefit Per Household. Retreived November 11, 2013, from http://www.fns.usda.gov/pd/19SNAPavg$HH.htm.
58
PLEASE DO NOT CITE FINDINGS FROM THIS REPORT WITHOUT PERMISSION OF THE AUTHORS
Widom, Rebecca, and Olivia Arvizú Martinez. 2007. “Keeping Food on the Table: Challenges to Food Stamps Retention in New York City.” New York, NY: Urban Justice Center, Homelessness Outreach and Prevention Project.
Yen, Steven T., Margaret Andrews, Zhuo Chen, and David B. Eastwood. 2008. “Food stamp participation and food insecurity: An instrumental variables approach.” American Journal of Agricultural Economics, 90(1): 117–32.
Zenk, Shannon N. Amy J. Schulz, Barbara A. Isreal, Sherman A. James, Shuming Bao and Mark L. Wilson. 2005. “Neighborhood Racial Composition, Neighborhood Poverty, and the Spatial Accessibility of Supermarkets in Metropolitan Detroit.” American Journal of Public Health 95(4): 660–67.
59